A, The scale represents false discovery rate–corrected P values for the statistical significance of the correlations of cumulative prenatal exposure to each drug with measures of local volume at each point on the surface of the brain in statistical models that included newborns separately in each drug-exposed group and their matched unexposed controls. B, Association of cumulative tobacco and alcohol exposure across all newborns in the cohort (N = 118). C, False discovery rate–corrected statistical maps correlating cumulative drug exposure with local volumes in a linear regression model including all newborns in the drug-exposed group but excluding the unexposed controls. ACA indicates anterior cingulate cortex; FG, fusiform gyrus; GR, gyrus rectus; IFG, inferior frontal gyrus; IOG, inferior occipital gyrus; IPL, inferior parietal lobule; ITG, inferior temporal gyrus; MFG, middle frontal gyrus; MOG, middle occipital gyrus; MTG, middle temporal gyrus; OFG, orbitofrontal gyrus; PCC, posterior cingulate cortex; PHG, parahippocampal gyrus; SPL, superior parietal lobule; PrPo, pre- and postcentral gyrus; SFG, superior frontal gyrus; SG, supramarginal gyrus; STG, superior temporal gyrus.
The scale represents false discovery rate–corrected P values for the statistical significance of the correlations of cumulative prenatal exposure to each drug with brain imaging measures at each voxel. The model was applied separately to each exposure group with its matched unexposed controls. Significant correlations of cumulative exposure with each imaging measure were similar in location across exposures. aCR indicates anterior corona radiata; aCS, anterior centrum semiovale; ALIC, anterior limb of the internal capsule; BG, basal ganglia; EC, external capsule; NAA, N-acetyl aspartate; pCR, posterior corona radiata; pCS, posterior centrum semiovale; PLIC, posterior limb of the internal capsule; Th, thalamus.
The color scale represents false discovery rate–corrected P values for the statistical significance of the voxelwise correlations of age with each imaging measure. Sex was included as a covariate. A, Correlation of age with local volumes when brains were not rescaled to a common volume. Nearly all regions correlated positively with age; thus, we inferred that all regions were growing through this period of development (postmenstrual ages of 37-46 weeks). Exceptions included regions that did not vary significantly with age (small portions of the inferior frontal gyri [IFG], superior frontal gyri [SFG], and orbital frontal gyri [OFG]) or that declined with age (parahippocampal gyrus [PHG]). B, Correlation of age with local volumes are shown for brains when rescaled to a common volume. We inferred that regions growing more slowly than overall brain volume included those at the dorsal (pre- and postcentral gyrus [PrPo], SFG), inferior (orbitofrontal gyrus [OFG] and PHG), and medial (SFG) brain surfaces, and that regions growing faster than overall brain volume included those over the lateral surface of the temporal (superior temporal gyrus [STG], middle temporal gyrus, [MTG], and inferior temporal gyrus [ITG]) and parietal (inferior parietal lobe [IPL]) lobes, as well as the medial posterior (posterior cingulate cortex [PCC]) and orbital (gyrus rectus [GR]) surfaces. Note that the prenatal exposure associations in Figure 1 map onto these age correlations, suggesting that the exposures are associated with accelerated growth of regions that normally grow quickly and with slowed growth of regions that normally grow slowly compared with overall brain volume. ADC indicates average diffusion coefficient; FA, fractional anisotropy; NAA, N-acetyl aspartate.
eMethods. MRI Pulse Sequences, Image Processing Methods, Post Hoc Statistical Analyses
eResults. Additional Maternal Characteristics
eDiscussion. Neurochemical Action of Exposure, Drugs Relation to Prior Studies, and Additional Limitations
eTable. Infant Outcomes at Postnatal Age 12 Months
eFigure 1. Examples of MPCSI Data Quality, T2 Relaxometry Image Quality, DTI Data Quality: ADC Map, and DTI Data Quality: Fiber Direction Map
eFigure 2. Anatomical MRI: Continuous Exposure Measures, with Rescaling
eFigure 3. Anatomical MRI: Dichotomous Exposure, without Rescaling
eFigure 4. Anatomical MRI: Modeling of Co-Occurring Exposures, Continuous Exposure Measures, Without Rescaling
eFigure 5. Anatomical MRI: Modeling of Co-Occurring Exposures, Continuous Exposure Measures, with Rescaling
eFigure 6. Anatomical MRI: Modeling of Co-Occurring Exposures, Dichotomous Exposure, Without Rescaling
eFigure 7. Anatomical MRI: Modeling of Co-Occurring Exposures, Dichotomous Exposure, with Rescaling
eFigure 8. Dichotomous Exposure and Dose-Response Association: Fractional Anisotropy
eFigure 9. Dichotomous Exposure and Dose-Response Association: Average Diffusion Coefficient
eFigure 10. Modeling of Co-Occurring Exposures, Continuous Exposure Measures: Fractional Anisotropy
eFigure 11. Modeling of Co-Occurring Exposures, Continuous Exposure Measures: Average Diffusion Coefficient
eFigure 12. Modeling of Co-Occurring Exposures, Continuous Exposure Measures: NAA
eFigure 13. Modeling of Co-Occurring Exposures, Continuous Exposure Measures: Choline
eFigure 14. Modeling of Co-Occurring Exposures, Continuous Exposure Measures: Creatine
eFigure 15. Mediation Model
eFigure 16. Anatomical MRI Mediation: Vineland Adaptive Behavior Scores
eFigure 17. Anatomical MRI Mediation: Vineland Communication Scores
eFigure 18. Anatomical MRI Mediation: Bayley Gross Motor
eFigure 19. DTI Mediation: Vineland Adaptive Behavior Scores
eFigure 20. DTI Mediation: Vineland Living Skill Scores
eFigure 21. DTI Mediation: Bayley Receptive Communication Scores
eFigure 22. Scatterplots for Exposure Associations in Matched Groups: Anatomical MRI
eFigure 23. Scatterplots for Exposure Associations in Matched Groups: DTI
eFigure 24. Scatterplots for Exposure Associations in Matched Groups: Relaxometry & MRSI
eFigure 25. Scatterplots for Dose-Response Associations: Anatomical MRI
eFigure 26. Scatterplots for Inter-Modality Correlations
Customize your JAMA Network experience by selecting one or more topics from the list below.
Peterson BS, Rosen T, Dingman S, et al. Associations of Maternal Prenatal Drug Abuse With Measures of Newborn Brain Structure, Tissue Organization, and Metabolite Concentrations. JAMA Pediatr. 2020;174(9):831–842. doi:10.1001/jamapediatrics.2020.1622
What are the associations of prenatal illicit drug exposure to marijuana, cocaine, and methadone and/or heroin with measures of brain organization in newborns?
In this cohort study of 118 economically disadvantaged mothers and their newborns, prenatal exposure to marijuana, cocaine, or opioids was associated with measures of brain structure, tissue organization, and metabolite concentrations; the associations were similar to those of advancing age with brain measures in unexposed newborns.
Prenatal drug exposure to marijuana, cocaine, and opioids is associated with measures of newborn brain tissue in patterns that suggest an exaggeration of normal fetal brain maturation.
Increasing rates of illicit drug use during pregnancy may be associated with risk for long-term health problems in prenatally exposed children.
To identify the associations of prenatal exposure to illicit drugs with organization of the newborn brain.
Design, Setting, and Participants
For this cohort study, a volunteer sample of 210 illicit drug–using and nonusing mothers and their newborns was enrolled from prenatal clinics and drug abuse treatment programs in New York, New York. Enrollment, scanning, and long-term follow-up occurred from September 2004 through February 2012, and image processing and statistical analyses continued through fall 2018. In addition to 26 participants with incomplete data, a total of 64 mothers were lost to follow-up during pregnancy, and 13 newborns were lost to follow-up at birth because of perinatal complications.
Newborns were assigned to 1 of 4 primary exposure groups based on the history of most frequent maternal drug use: marijuana, cocaine, methadone maintenance, and/or heroin. Unexposed newborns were controls.
Main Outcomes and Measures
Unsedated magnetic resonance imaging (MRI) of newborn brains was performed shortly after birth. Infant neurodevelopmental outcomes were assessed at age 12 months. MRI modalities included anatomical imaging, diffusion tensor imaging, T2 relaxometry, and magnetic resonance spectroscopic imaging. Infant neurodevelopmental outcomes included Bayley scales of infant development–III and Vineland Adaptive Behavior Scales. Statistical analyses were performed with results represented on the brain images.
Of 118 mothers, 42 (35%) were in the control group (mean [SD] age, 25.9 [6.1] years), 29 (25%) were in the cocaine group (mean [SD] age, 29.0 [6.1] years), 29 (25%) were in the marijuana group (mean [SD] age, 24.3 [5.5] years), and 18 (15%) were in the methadone and/or heroin group (mean [SD] age, 30.9 [5.7] years). Not all newborns could be scanned successfully; therefore, usable MRIs were acquired for 118 newborns from predominantly minority groups and with economically disadvantaged mothers. Anatomic abnormalities were detected in similar locations across all 3 drug exposures and included smaller volumes in the dorsal, medial, and ventral surfaces of the frontal lobe and dose-related increases in volumes in the lateral temporal lobe, dorsal parietal lobe, and superior frontal gyrus. Dose-related increases in diffusion tensor measures of tissue organization, decreases in T2 relaxometry times, and increases in spectroscopy metabolite concentrations were similar across exposures. These associations of exposures with brain measures were similar to the associations of newborn age with brain measures. The anatomic and diffusion tensor imaging measures suppressively mediated the associations of prenatal exposure with poorer 12-month infant outcomes.
Conclusions and Relevance
The findings suggest that prenatal drug exposure is associated with measures of newborn brain tissue in patterns that may indicate that exposures accelerated normal fetal brain maturation, which in turn mediated the associations with poorer 12-month infant outcomes.
Substance use during pregnancy has become commonplace and may be associated with risk for long-term cognitive, behavioral, and emotional health problems in prenatally exposed children. More than 60% of pregnant women report using illegal substances, and 15% have a substance use disorder1,2; these women are often younger and are experiencing material and social deprivation.3 Owing partially to its decriminalization, marijuana use during pregnancy has increased nearly 7-fold in the past decade.4-6 The US opioid crisis has been associated with a concomitant epidemic of use among pregnant women and with withdrawal syndromes in their prenatally exposed newborns.7-9 Cocaine use remains relatively stable, with approximately 750 000 pregnant US women using this drug annually.10
Long-term outcome studies face the challenge of disentangling the effects of prenatal illicit drug exposure from those of co-occurring substance use, poverty, maternal stress, and poor prenatal care. Nevertheless, consensus is emerging that prenatal cocaine exposure is associated with preterm birth,11-13 sustained newborn jitteriness,11,13-17 slowed fetal and postnatal somatic growth,18-20 delayed language development,21-23 impaired self-regulation,24 and poor sustained attention.25-27 Prenatal marijuana exposure slows somatic growth28; impairs memory, verbal reasoning, and attention; and increases impulsivity, hyperactivity, and oppositional behaviors.29-33 Prenatal methadone exposure produces motor and cognitive impairments,34,35 inattention, impulsivity, and hyperactivity.36-38 Prenatal alcohol and tobacco exposure can lead to similar problems in childhood.39 Thus, although differing exposures are associated with unique developmental problems, they have in common an association with the development of inattention, hyperactivity, and impulsivity in later childhood.39,40
This study evaluated the associations of prenatal exposure to marijuana, opioids, or cocaine with magnetic resonance imaging (MRI) measures in the neonatal brain as soon after exposure as possible. Given the common associations that prenatal cocaine, marijuana, and opioid exposure have with reducing brain growth and increasing attention-deficit/hyperactivity disorder–like symptoms, we hypothesized that we would detect exposure-associated reductions in brain volumes in anatomic MRI scans localized to the frontal lobe and other higher-order association areas. Other MRI modalities were acquired for post hoc analyses to help identify the likely cellular bases for volume reductions.
For this cohort study, illicit drug–using pregnant women were recruited from prenatal clinics and drug abuse treatment programs in Manhattan, New York City. Healthy pregnant women were recruited from prenatal clinics at New York Presbyterian Hospital, New York, New York. All women provided informed written consent for participation. Procedures were approved by the institutional review board of the New York State Psychiatric Institute; participants were protected by a certificate of confidentiality. Mothers were paid for participation in gift cards for newborn supplies. Statistical analyses were performed with results represented on the brain images.
Inclusion criteria consisted of (1) maternal age of 18 to 45 years; (2) maternal use of marijuana, cocaine, or methadone and/or heroin for the exposed groups or no history of illicit drug or alcohol use during pregnancy for the control group; and (3) newborn gestational age at birth of at least 37 weeks.
Exclusion criteria consisted of (1) maternal HIV infection or AIDS, seizure disorder, malignant tumors, diabetes, or use of psychoactive medications; (2) infant congenital anomalies, genetic syndromes, or significant neonatal complications; (3) maternal alcohol use of more than 2 ounces per day for the exposed group or more than 1 ounce per day in any trimester for the control group; and (4) urine toxicology or self-reports indicating use of other illicit substances.
Illicit drug use history was obtained using drug use questions from the Addiction Severity Index41-43 administered at enrollment, during each subsequent trimester, and at birth. Drug use was confirmed with random urine toxicology screens for the mother during pregnancy and for the newborn at delivery and through maternal and medical record reviews. Exposed newborns were divided into 3 groups based on most frequent maternal drug use: marijuana, cocaine, or methadone maintenance therapy for narcotic addiction with or without concurrent use of heroin (methadone and/or heroin). Every effort was made to recruit mothers who were using only a single drug.
We enrolled 210 mother and newborn pairs: controls (n = 63), cocaine users (n = 53), marijuana users (n = 55), and methadone and/or heroin users (n = 39). During pregnancy, 64 of 210 (30.5%) were lost to follow-up because of disinterest, maternal incarceration, changes in residence, or spontaneous abortions. At birth, 13 of the remaining 146 newborns (8.9%) were lost to follow-up owing to prematurity (methadone and/or heroin [n = 2], cocaine [n = 1]), major congenital heart or pulmonary disease (cocaine [n = 1], methadone and/or heroin [n = 1], and marijuana [n = 1]), abnormal anatomic scan findings reflecting infarction (cocaine [n = 1], marijuana [n = 1]), or foster care (cocaine [n = 2], methadone and/or heroin [n = 2], and marijuana [n = 1]).
We acquired anatomical MRIs to study brain structure, diffusion tensor imaging and T2 time relaxometry to study brain tissue organization, and magnetic resonance spectroscopic imaging to study chemical metabolites. MRIs were obtained for 133 unsedated newborns at 37 to 46 weeks’ postmenstrual age.
Usable MRI data in at least 1 modality were obtained in 118 newborns: marijuana (n = 29), cocaine (n = 29), methadone and/or heroin (n = 18), and unexposed controls (n = 42). Not all MRI modalities could be acquired for every newborn, leaving a subset with usable MRIs in each modality. We group matched each drug-exposed group to a randomly sampled subset of the unexposed newborns, with the constraint that groups did not differ significantly by sex or postmenstrual age at the time of MRI, yielding the following: (1) anatomic MRI: marijuana (n = 23) and controls (n = 37), cocaine (n = 25) and controls (n = 37), methadone and/or heroin (n = 17) and controls (n = 33); (2) diffusion tensor imaging: marijuana (n = 20) and controls (n = 26), cocaine (n = 16) and controls (n = 21), methadone and/or heroin (n = 13) and controls (n = 20); (3) T2 relaxometry: cocaine (n = 17) and controls (n = 20), marijuana (n = 14) and controls (n = 18), methadone and/or heroin (n = 11) and controls (n = 18); and (4) 1H-multi-planar chemical shift imaging44: marijuana (n = 22) and controls (n = 29), cocaine (n = 18) and controls (n = 29), and methadone and/or heroin (n = 13) and controls (n = 29). Enrollment, scanning, and long-term follow-up occurred from September 2004 through February 2012, and image processing and statistical analyses continued through fall 2018.
English- or Spanish-language interviews were completed to collect maternal demographics, psychiatric history,45 socioeconomic status,46 depression severity,47 and anxiety severity.48 Maternal IQ was assessed prenatally.49 Newborn assessments at age 12 months included Bayley Scales of Infant Development–III50 and Vineland Adaptive Behavior Scales.51 Race/ethnicity was self-reported. MRI, pulse sequences, and image processing procedures are detailed in the eMethods in the Supplement.
We applied multiple linear regression at each voxel in template space within each MRI modality to test our a priori hypothesis that prenatal drug exposure would be associated with alterations in the newborn brain. The general model for testing exposure correlates in each modality was
Imaging Measure = β0 + (β1 × Drug Exposure) + (β2 × Alcohol) + (β3 × Tobacco) + (β4 × Postmenstrual Age) + (β5 × Sex) + ϵ.
The dependent measure for anatomical analyses was the signed euclidean distance at each point on the surface of each newborn brain from the corresponding point on the surface of the template, which we designate for simplicity here as a measure of local volume. For diffusion tensor imaging, the dependent measure was either fractional anisotropy (FA) or average diffusion coefficient (ADC), and for relaxometry, it was T2 relaxation time; for magnetic resonance spectroscopic imaging, it was N-acetylaspartate (NAA) concentration. The independent variable (drug exposure) was prenatal exposure to a particular drug in the specific exposure group and its matched unexposed controls, with exposure coded as a continuous variable representing cumulative exposure throughout pregnancy based on maternal reports of daily use, converted to either grams of cocaine, joints of marijuana, grams of methadone, cigarette packs, or ounces of alcohol (Table 1).52,53 Cumulative exposure distributions were highly skewed, with most values being small but some large. Preliminary analyses showed that drug exposure had logarithmic rather than linear associations with brain measures. We considered various log transformations, but the one that could best address 0 values was the inverse hyperbolic sine (IHS)54,55:
where the transformed value approaches the log of twice the exposure for large values and is 0 for no exposure. Covariates in all analyses were newborn postmenstrual age at MRI, newborn sex, and cumulative maternal tobacco and alcohol use during pregnancy. We also assessed the effects of including as additional covariates in our models maternal age, socioeconomic status, race/ethnicity, depression severity, anxiety severity, or prenatal stress.
P values in each statistical map within each imaging modality were corrected for the number of statistical comparisons. We applied the Benjamini–Yekutieli procedure56 for false discovery rate to control for the expected number of false-positive findings. False discovery rate–corrected, 2-sided P values were color coded and displayed on the template brain.
We assessed dose-response associations by correlating cumulative drug use with imaging measures within each exposure group excluding the unexposed controls and covarying for postmenopausal age and sex.
Testing brain-based mediation of the associations of exposures with 12-month infant outcomes was performed voxelwise using the Sobel test (eFigure 15 in the Supplement), and we assessed whether mediation was complementary or suppressive (eMethods and eFigure 15 in the Supplement).57,58
Of 118 mothers whose newborns had usable MRI data, 42 (35%) were in the control group (mean [SD] age, 25.9 [6.1] years), 29 (25%) were in the cocaine group (29.0 [6.1] years), 29 (25%) were in the marijuana group (24.3 [5.5] years), and 18 (15%) were in the methadone and/or heroin group (30.9 [5.7] years). The 133 mothers with newborns who underwent MRI did not differ significantly in demographics from the 77 enrolled mothers with newborns who did not undergo MRI (eResults in the Supplement). Among the 118 mothers of newborns with usable MRI scans, polydrug use was present in 4 of 29 (13.8%) in the cocaine group, 1 of 29 (3.5%) in the marijuana group, and 6 of 18 (33.3%) in the methadone and/or heroin group (Table 1). By design, reported alcohol use was minimal in each exposure group for the duration of pregnancy. The number of women who smoked more than half a pack per day of cigarettes was 6 (20.7%) in the cocaine group, 2 (11.1%) in the marijuana group, and 6 (33.3%) in the methadone and/or heroin group (Table 1). Additional maternal and newborn characteristics are shown in eFigure 12 and the eResults in the Supplement.
MRI data quality is shown in eFigure 1 in the Supplement. Compared with their matched, unexposed controls, all 3 exposure groups had dose-related volume reductions in the same general locations, including dorsal and lateral surfaces of the frontal lobe and mesial and inferior cerebral surfaces. The cocaine and marijuana groups also had smaller volumes in most of the lateral surface of the temporal lobe, extending to the inferolateral occipital surface (Figure 1). Reduced volumes associated with co-occurring alcohol use were present in similar locations although smaller in spatial extent (Figure 1).
Prenatal exposure was also associated with regional enlargement, particularly of the lateral temporal and inferolateral frontal lobes and mesial wall in the methadone and/or heroin group, with smaller foci in similar locations in the cocaine group (Figure 1).
Findings were similar when assessing exposure correlates for all newborns entered into a single exposure model and when treating individual drug exposures as dichotomous variables for any exposure during pregnancy (eFigures 3-7 in the Supplement). Regional volume reductions exhibited linear associations with IHS-transformed measures of cumulative exposure during pregnancy within each exposure group (eFigure 22 in the Supplement).
Prenatal cocaine, marijuana, and methadone and/or heroin exposure were all significantly associated with increased FA (Figure 2) and reduced ADC (Figure 2) throughout frontal and parietal white matter. Marijuana was additionally associated with increased FA in the anterior limb of the internal capsule and marijuana and cocaine with reduced FA in the posterior limb of the internal capsule. Scattered significant voxels at the interfaces of gray matter, white matter, and cerebrospinal fluid may represent partial volume effects, but large clusters in regions in deep white matter distant from tissue boundaries likely represent findings in only white matter.
All 3 exposures were associated with reduced T2 relaxation times in frontal and parietal white matter; cocaine and methadone and/or heroin were associated with increased T2 times in ventral striatum and thalamus (Figure 2). Cocaine and marijuana exposure but not methadone and/or heroin were associated with increased NAA in deep white matter of the frontal and parietal lobes (Figure 2). Post hoc analyses showed that choline and creatine concentrations exhibited similar associations as NAA (eFigures 12-14 in the Supplement).
Advancing newborn age was associated with smaller volumes of frontal, posterior, and inferior brain surfaces when controlling for overall brain size, suggesting slower growth compared with overall brain growth. Newborn age was also associated with widespread increases in FA and reductions in ADC and T2 times throughout the brain, and with increased NAA concentrations in basal ganglia and thalamus (Figure 3).
Findings were similar when maternal age, socioeconomic status, race/ethnicity, depression severity, anxiety severity, or prenatal stress were included as additional covariates in our statistical models.
Twelve-month Bayley and Vineland scores were lower for exposed infants (eTable in the Supplement). Exposure-associated reductions in dorsal, inferior, and medial frontal volumes partially mediated the associations of any prenatal drug exposure with 12-month Vineland measures of adaptive behavior (eFigure 16 in the Supplement) and communication (eFigure 17 in the Supplement), and reduced volumes in premotor regions partially mediated exposure associations with lower Bayley gross motor scores (eFigure 18 in the Supplement). In all regions, mediation was suppressive rather than complementary (eFigures 16-18 in the Supplement).57,58 Regional enlargements did not significantly mediate outcomes. Lower mean diffusivity measures partially mediated exposure associations with lower scores for Vineland adaptive behavior and living skills and Bayley receptive communication (eFigures 19-21 in the Supplement).
We assessed the intercorrelations of imaging measures across MRI modalities for all available newborns in regions exhibiting a commonality of findings across exposures. Scatterplots for imaging measures adjusted for newborn postmenstrual age and sex showed significant linear correlations between measures in all MRI modalities (eFigure 26 in the Supplement).
In this study, we detected significant associations of prenatal illicit drug exposure with measures of brain structure, tissue organization, and metabolites in newborns. Overall head circumference and brain volume were smaller in drug-exposed newborns (significantly in the methadone and/or heroin group) when controlling for length as a scaling effect and postmenstrual age as a maturational index (Table 2) and thus exceeded the anticipated association between illicit drug use and slowing of overall somatic growth.18-20,28,39 Smaller overall brain volumes derived from smaller regions in similar locations across all 3 exposures, including dorsal, medial, and ventral surfaces of the frontal lobe (Figure 1). These findings were present when comparing newborns in individual exposure groups with unexposed controls, when controlling for co-occurring exposures in a single exposure model that included all newborns, and when treating individual drug exposures as either dichotomous variables (eFigures 6-9 in the Supplement) or continuous measures of cumulative exposure (eFigures 2-7 and 10-14 in the Supplement). Each exposure was also associated with regional enlargements (lateral temporal lobe, dorsal parietal lobe, and superior frontal gyrus) in direct proportion to cumulative exposure measures and in similar locations across exposures, reaching statistical significance for cocaine and methadone and/or heroin (Figure 1), especially when adjusting for overall brain volume (eFigures 2, 5, and 7 in the Supplement).
These common associations of prenatal exposures with brain structure are similar in direction and location to the overall correlates of advancing age on brain structure, which included slower growth of frontal, posterior, and inferior brain surfaces and more rapid growth of lateral temporal surfaces compared with overall brain growth (Figure 3). Prenatal drug exposure therefore seemed to be associated with slowed brain growth disproportionately in regions that grow more slowly and with accelerated growth in regions that grow more quickly. Accelerated growth in lateral temporal cortices was associated with cumulative exposure (eFigure 25 in the Supplement). Because most drug exposure occurred in the first trimester, however, dose-related acceleration in growth could instead represent a timing-specific consequence of exposure early in gestation. Regions of exaggerated slowing of growth were not associated with cumulative dose and may be independent of both exposure timing and dose.
Developmental interpretations of our anatomical findings are supported by findings in each of the other modalities. Advancing age was associated with widespread increases in FA and reductions in ADC and T2 times in our data and previous studies59-61 and in our findings of exposure-related increases in NAA concentrations in subcortical white and gray matter (Figure 2). All 3 exposures were associated, to varying degrees, with abnormalities in these same regions and in the same direction, supporting interpretation of the anatomic findings that exposures may be associated with accelerated fetal brain maturation. Each exposure, for example, was associated with increased FA and reduced ADC (Figure 2), indicating a dose-related greater restriction on the diffusion of water perpendicular to the direction of axons within white matter, most likely from more myelin or greater axon density62 throughout frontal and parietal white matter and in proportion to cumulative exposure (eFigure 23 in the Supplement). The dose-related decrease in T2 relaxation times in these same regions for each exposure (Figure 2) was also consistent with more myelin or greater cell density because both reduce brain free water content and thereby reduce T2 times with advancing age in early postnatal life.60,63,64 Finally, previous studies have reported that NAA concentrations and axon density increase rapidly in white matter after birth.65-68 In our data, higher NAA concentrations were associated with newborn age and with prenatal cocaine and marijuana exposure in large white matter fiber bundles and basal ganglia (Figure 2 and eFigure 12 in the Supplement), suggesting that increased density or metabolism of neurons69,70 accompanies advancing newborn age and higher exposures.
Regions of slow anatomic maturation suppressively mediated lower infant 12-month Vineland social communication and adaptive behavior scores (eFigures 16 and 17 in the Supplement). Significant suppressive mediation indicates that slowed regional maturation was likely an adaptive response to prenatal illicit drug exposure, attenuating what would otherwise have been worse 12-month outcomes.57,58 We speculate that slowed maturation in dorsal and inferior frontal brain regions is a response to the deleterious consequences of illicit drug exposure elsewhere in the brain.
Although the specific biochemical consequences of each drug would indicate specific brain effects (eDiscussion in the Supplement),39,71 our findings instead suggest mostly a commonality of drug effects, presumably operating through a final common pathway. The commonality of drug effects is underscored by similar findings across exposures in each MRI modality and by the significant correlations of imaging measures across modalities (eFigure 25 in the Supplement) showing that newborns who had the largest exposure correlates in one modality also had the largest exposure correlates in another modality. Preclinical studies (eDiscussion in the Supplement) suggest that the reductions in neuronal growth, differentiation, dendritic complexity, synaptic density, and neurotransmitter signaling that all 3 illicit drugs produce may be associated with slowed development in regions where growth is normally slower than in the rest of the brain. Regional accelerations in growth and increased density or myelination of white matter, in contrast, may be associated with general drug effects on increasing cellular excitation, which would in turn influence the activity-dependent survival of axons, dendrites, and synapses66 and proliferation of oligodendrocyte precursors72-75 during circuit formation. Activity-dependent processes would account for the dose-dependent increases in volumes of the lateral temporal, posterior parietal, and dorsal medial surfaces (Figure 1).
This study has limitations. Co-exposure is a possible reason for the commonality of findings across exposure groups, particularly because women with substance abuse typically use a cocktail of drugs, making differentiation of specific effects difficult even with very large samples. Nevertheless, we strove to exclude women using illicit drugs other than those in the exposure groups, rates of co-occurring drug use were relatively low (Table 1), and findings persisted when all newborns were entered into a single model for all concurrent exposures. The most common co-exposures were tobacco or alcohol use, and all statistical analyses covaried for them. Moreover, alcohol use was minimal and tobacco use was limited in each exposure group (Table 1). We also modeled in each MRI modality the correlates of drug, alcohol, and tobacco use (Figure 1) (eFigures 4-7 and 10-14 in the Supplement). Tobacco correlates were dissimilar to the correlates of the other exposures; alcohol correlates included some regions of reduced volume in anatomic images overlapping with drug associations but did not show the regional enlargements associated with drug exposures. Furthermore, the associations of tobacco and alcohol with FA, ADC, and NAA measures were few and were not similar to the drug correlates (Figure 1). In addition, we demonstrated a dose-response association of drug exposures with brain measures, even when controlling for co-occurring tobacco and alcohol exposures, when using continuous measures of drug exposure in models for the correlates of co-occurring exposures and in scatterplots displaying the dose-response associations (eFigures 22-25 in the Supplement).
We also cannot exclude the possibility that other potential confounders, such as socioeconomic status, prenatal nutrition, or stress, contributed to the commonality of findings, especially because the sample size limited our ability to model confounder effects. However, covarying for these factors had negligible effects on our findings. In addition, the demographics of the exposure and control groups were similar and controlled for many of these potential confounders.
Sample sizes for individual exposure were modest, limiting our statistical power to discern associations. Moreover, study participation, sample size, and eligibility criteria limit the generalizability of our findings. Many mothers discontinued illicit drug use early in their pregnancies, limiting exposures in the second and third trimesters and the brain correlates specific to those periods of fetal development. Unassessed postnatal exposures could have contributed to poorer 12-month neurodevelopmental outcomes in the drug-exposed groups but would not have contributed to neonatal MRI measures. Furthermore, inconsistent follow-up assessments limited our analyses of brain-based mediators of the associations of exposure with long-term outcomes, which should be regarded as heuristic rather than conclusive. In addition, inferring developmental correlates from cross-sectional data such as ours can be hazardous; our inference that exposure associations represent an exaggeration of normal brain maturation requires confirmation with repeated scanning in a long-term study that defines growth trajectories in the newborn brain.
The findings suggest that prenatal illicit drug exposure is associated with measures of newborn brain tissue and that these associations may represent an exaggeration of normal fetal brain maturation, which in turn may mediate the association of prenatal exposures with poorer 12-month infant outcomes.
Accepted for Publication: April 15, 2020.
Corresponding Author: Bradley S. Peterson, MD, Department of Pediatrics, Children’s Hospital Los Angeles, 4650 Sunset Blvd, Mail Stop #135, Los Angeles, CA 90027 (email@example.com).
Published Online: June 15, 2020. doi:10.1001/jamapediatrics.2020.1622
Author Contributions: Dr Peterson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: B.S. Peterson, Rosen, Dingman, Branch.
Acquisition, analysis, or interpretation of data: B.S. Peterson, Dingman, Toth, Sawardekar, Hao, Liu, Xu, Dong, J.B. Peterson, Ryoo, Serino, Branch, Bansal.
Drafting of the manuscript: B.S. Peterson, Toth, Ryoo, Branch, Bansal.
Critical revision of the manuscript for important intellectual content: B.S. Peterson, Rosen, Dingman, Sawardekar, Hao, Liu, Xu, Dong, J.B. Peterson, Serino, Bansal.
Statistical analysis: B.S. Peterson, Sawardekar, Hao, Xu, Ryoo, Bansal.
Obtained funding: B.S. Peterson, Dingman.
Administrative, technical, or material support: B.S. Peterson, Rosen, Dingman, Toth, Sawardekar, Xu, J.B. Peterson, Branch.
Supervision: B.S. Peterson, Dingman, Xu, Bansal.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by grant R01 DA017820 (Dr Peterson) from the National Institutes of Health and funding from Children’s Hospital Los Angeles and the University of Southern California.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Information: This research was made possible by provision of data by New York State Psychiatric Institute and Columbia University, New York, New York.