• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Neurosci. Author manuscript; available in PMC Sep 17, 2010.
Published in final edited form as:
PMCID: PMC2847574

Differentiating prenatal exposure to methamphetamine and alcohol versus alcohol and not methamphetamine using tensor based brain morphometry and discriminant analysis

Elizabeth R. Sowell, Ph.D.,1,2 Alex D. Leow, M.D., Ph.D.,1 Susan Y. Bookheimer, Ph.D.,3 Lynne M. Smith, M.D.,4 Mary J. O’Connor, Ph.D.,3 Eric Kan,1,2 Carly Rosso,1,2 Suzanne Houston, M.A.,1,2 Ivo D. Dinov, Ph.D.,1 and Paul M. Thompson, Ph.D.1


Here we investigate the effects of prenatal exposure to methamphetamine (MA) on local brain volume using magnetic resonance imaging. Because many who use MA during pregnancy also use alcohol, a known teratogen, we examined whether local brain volumes differed among 61 children (ages 5 to 15), 21 with prenatal MA exposure, 18 with concomitant prenatal alcohol exposure (the MAA group), 13 with heavy prenatal alcohol but not MA exposure (ALC group), and 27 unexposed controls (CON group). Volume reductions were observed in both exposure groups relative to controls in striatal and thalamic regions bilaterally, and right prefrontal and left occipitoparietal cortices. Striatal volume reductions were more severe in the MAA group than in the ALC group, and within the MAA group, a negative correlation between full-scale IQ (FSIQ) scores and caudate volume was observed. Limbic structures including the anterior and posterior cingulate, the inferior frontal gyrus (IFG) and ventral and lateral temporal lobes bilaterally were increased in volume in both exposure groups. Further, cingulate and right IFG volume increases were more pronounced in the MAA than ALC group. Discriminant function analyses using local volume measurements and FSIQ were used to predict group membership, yielding factor scores that correctly classified 72% of participants in jackknife analyses. These findings suggest that striatal and limbic structures, known to be sites of neurotoxicity in adult MA abusers, may be more vulnerable to prenatal MA exposure than alcohol exposure, and that more severe striatal damage is associated with more severe cognitive deficit.

Keywords: Teratogen, cognitive, development, imaging, neurobehavioral, Methamphetamine

MA abuse is a continuing problem, and data from the 2002–2004 National Surveys on Drug Use and Health indicated 16–17 million Americans over the age of 12 have used methamphetamine (Colliver et al., 2006). Of those, approximately 19,000 were pregnant women (Colliver et al., 2006). Significant advances have been made in understanding the neurotoxic effects of MA on brain structure in adults (reviewed in (Chang et al., 2007; Berman et al., 2008)). Abuse of the drug over time causes damage to dopaminergic and serotonergic brain regions, most prominent in the basal ganglia. Limbic volume reductions have been observed in the cingulate and medial temporal lobes in MA abusers (Thompson et al., 2004). The effects of prenatal exposure to MA, when the brain is still developing, is not yet well understood.

Brain proton spectroscopy studies revealed higher creatine in the striatum and frontal lobe white matter of children with prenatal MA exposure, reflecting MA-induced changes in cellular metabolism and perhaps accelerated neuronal and glial development (Smith et al., 2001; Chang et al., 2009). Quantitative morphological analysis revealed smaller putamen, globus pallidus, and hippocampal volumes in children with prenatal MA exposure. Volume reduction in these structures correlated with poorer performance on attention and verbal memory (Chang et al., 2004). Diffusion tensor imaging studies of children with prenatal MA exposure revealed abnormalities of frontal and parietal white matter including increased fractional anisotropy (FA) and decreased diffusion (Cloak et al., 2009). We have also seen increased FA in frontal white matter in some of the same subjects studied here (Colby et al., Submitted). Finally, recent work by our group revealed more diffuse brain activation in children with prenatal MA exposure during verbal learning (Lu et al., 2009). Taken together, the small brain imaging literature on prenatal MA exposure shows prominent effects in the fronto-striatal network, known to be impacted from animal and human studies of adult MA abusers.

Efforts to understand specific effects of prenatal methamphetamine exposure on brain development in humans are complicated by high rates of concomitant alcohol use during pregnancy. Of pregnant MA users who participated in a study on prenatal MA exposure after giving birth, 49% reported having used alcohol during pregnancy (Smith et al., 2006). Alcohol is a known teratogen, frequently resulting in brain and cognitive abnormalities (reviewed in (Riley and McGee, 2005)). We evaluated the specific effects of prenatal exposure to MA on brain development by including a group of children with heavy prenatal alcohol and not MA exposure as a contrast group. Given findings from previous studies of MA exposure (Chang et al., 2004), and the known vulnerability of dopamine receptors to MA, we expected striatal volume reductions in our MA group (including those with MA and alcohol exposure; MAA) relative to both typically developing controls and to the alcohol-only contrast group. We expected cortical and white matter abnormalities to localize in frontal and parietal regions, given findings of white matter abnormalities in these regions from DTI studies (Cloak et al., 2009; Colby et al., Submitted).



Sixty-one participants were assigned to either a methamphetamine-exposed group (MAA), an alcohol-exposed group (ALC), or a non-exposed control group (CON) based on exposure histories, which were established by an extensive interview administered to the adult guardians of minor participants. Exact quantities and patterns of maternal use of alcohol and methamphetamine are difficult to ascertain in retrospective studies, so social, medical and/or legal records were used when available to confirm exposure histories. Of the 3 biological MA and/or alcohol-abusing mothers interviewed, all reported using MA and/or alcohol throughout pregnancy. The rest suffered legal and social problems as a result of their MA and/or alcohol use to the extent that their children were removed from their custody due to drug abuse in the home, neglect, or domestic violence according to reports of adoptive parents, legal, social or medical records.

Primary recruitment mechanisms for methamphetamine-exposed participants included a rehabilitation program serving women with children born positive for methamphetamine, a social-skills group at UCLA for children diagnosed with fetal alcohol spectrum disorders (FASDs) and word-of-mouth advertising. Of those recruited, participants were included in the methamphetamine group (MAA) if prenatal exposure to methamphetamine was sufficiently confirmed by parental or guardian report or reliable collateral report during the history interview (n=21). Eighteen of the MAA children were also exposed to alcohol prenatally. Children were excluded from the MAA group if 1) they had prenatal exposure to cocaine or other opiates, 2) they were less than 5 years of age, 3) they had an IQ of less than 70, 4) they had suffered a head injury with loss of consciousness over 20 minutes, 5) they had a physical (e.g., hemiparesis), psychiatric, or developmental disability (e.g., autism) that would preclude participation, 6) they had other potential known causes of mental deficiency (e.g., chromosomal disorders), 7) significant maternal illness that has increased risk for fetal hypoxia (e.g., sickle cell disease) was reported, or 8) they had metallic implants in the body which posed a risk for MRI.

Alcohol-exposed participants (n=13) were recruited from the same UCLA social skills group, described above, that serves children with fetal alcohol spectrum disorders. Participants in this group met the following inclusion criteria: 1) had exposure to four or more drinks per occasion at least once per week or 14 drinks or more per week and 2) were not exposed to methamphetamine during gestation. Exclusionary criteria were the same as for the MAA group.

Details of diagnostic procedures for alcohol-related disorders within the MAA and ALC groups are described in another report (O’Connor et al., 2006). Briefly, an experienced clinician examined alcohol-exposed children using the Diagnostic Guide for Fetal Alcohol Syndrome (FAS) and Related Conditions (Astley, 2004). This system uses a 4-digit diagnostic code reflecting the magnitude of expression of four key diagnostic features of FAS: (1) growth deficiency; (2) the FAS facial phenotype, including short palpebral fissures, flat philtrum, and thin upper lip; (3) central nervous system dysfunction; and (4) gestational alcohol exposure. All of the children with FASD diagnoses had histories of heavy prenatal alcohol exposure confirmed by maternal report, reliable collateral report, and/or medical or legal records. Using this 4-digit code, children were diagnosed with FAS, partial FAS (PFAS) alcohol-related neurodevelopmental disorder (ARND) or sentinal features. Recent studies have shown brain morphological abnormalities to be more severe in children with the facial phenotype to obtain a diagnosis of FAS or PFAS relative to children with alcohol related disorders but not facial dysmorphology {Astley, 2009 #1561}. Five subjects in the ALC group, and 4 subjects in the MAA group were diagnosed with FAS or PFAS, and the rest had ARND or sentinal features.

Non-exposed control participants (n=27) were recruited from the same Los Angeles communities as both the MAA and ALC exposed participants, in addition to bulk mail advertisements, magazine advertisements, and flyers distributed to local libraries. Participants were excluded from the CON group if they had exposure to more than two drinks on any occasion or an average of one drink or more per week. Other exclusionary criteria were exactly the same as for the other two participant groups.


All participants and their parents gave informed assent/consent according to procedures approved by the UCLA Institutional Review Board.

Image Acquisition

High-resolution T1-weighted sagittal volumes were collected from a 1.5 Tesla (T) Siemens Sonata scanner (repetition time (TR), 1900 ms; echo time (TE), 4.38 ms; flip angle, 15°; matrix size, 256 × 256 × 160; field of view (FOV), 256 mm; voxel size, 1 × 1 × 1 mm; acquisition time, 8min 8s). Two to four acquisitions were acquired for each subject. Raters blind to subject age, sex and diagnosis evaluated image quality, and data from at least two acquisitions were averaged to enhance signal-to-noise ratio.

Neurocognitive Evaluations

Children underwent extensive individual neuropsychological testing. Included among tests administered was an abbreviated version of the Wechsler Intelligence Scale for Children, Fourth Edition (WISC-IV, (Wechsler, 2003)). A prorated full scale intelligence quotient (FSIQ) was derived from the prorated Verbal Comprehension Index (VCI) and Perceptual Reasoning Index (PRI). This method is described in the WISC-IV manual, and there is evidence that prorated VCI and PRI are each highly correlated (rs ≥ 0.90) with actual index scores (Glass et al., 2008). FSIQ was not available for 2 of the controls subjects (not enough subtests administered due to time restrictions, and one 5-year-old was too young for the WISC-IV), and one of the ALC subjects (who was unable to tolerate testing). Socioeconomic status was estimated based on annual family income obtained in a parent interview.

Image Analysis

TBM Overview

MRI data were analyzed with tensor-based morphometry (TBM), which allows investigations of regional differences in the volumes of brain substructures by globally aligning all brain images into a common brain template before applying localized deformations to adjust each subject’s anatomy to match the global group-average template. Detailed methods are described elsewhere (Leow et al., 2005; Leow et al., 2007). These methods were developed in-house and have been extensively tested and validated (Yanovsky et al., In Press). They have been used previously to study normal brain development (Hua et al., 2009), brain dysmorphology in Fragile × syndrome (Lee et al., 2007), Williams Syndrome (Chiang et al., 2007), autism (Brun et al., 2009), Alzheimer’s Disease and mild cognitive impairment (Hua et al., 2008b; Leow et al., 2009) and childhood-onset schizophrenia (Gogtay et al., 2008). Briefly, our implementation of TBM involves 1) nonlinearly deforming all images by a mutual information (MI)-based inverse-consistent elastic registration algorithm to match a preselected brain template. 2) The 3D Jacobian determinant maps of the deformation fields (mathematically defined by computing the determinant of the point-wise Jacobian operator applied to these deformations) are used to gauge the local volume differences between individual images and the template. The determinant maps can be statistically analyzed on a voxel level to identify group differences in brain structure, such as localized atrophy or tissue excess.

The MI-based inverse-consistent elastic registration algorithm utilizes mutual information as the image similarity measure (Wells et al., 1996), and employs an elastic regularization (computationally realized in the Fourier domain) on the deformation field. A multi-resolution scheme is used to successively compute coarse-to-fine deformations using Fast Fourier Transforms (FFTs) at increasing spatial resolutions. Numerical convergence is checked every 20 iterations, and is defined as the point at which MI failed to increase by 0.001 after the prior iterations. 300 iterations are computed at each FFT resolution before increasing the resolution by a factor of 2 in each dimension (with the time step decreased to one-tenth of its previous value).

Image pre-processing

The MR images for each individual go through a series of automated and manual processes. First, multiple images acquired for each subject are averaged to improve the signal-to-noise (SNR) ratio. Second, extra non-brain tissues (i.e. scalp, orbits, brain stem) are automatically removed from the image (Shattuck and Leahy, 2002). Manual edits by expert image analysts were then made to remove any errors made by the automated brain extraction tool. The skull-stripped images are then run through an automated intensity-correction program to correct any inhomogeneity artifacts introduced by the magnet (Sled et al., 1998).

Generation of a Mean Anatomical Template

Each individual’s pre-processed image data was then affinely co-registered (9-parameter transformation) to a high-resolution MRI brain-only image volume of one representative subject. This step adjusts for individual differences in global brain scale and head alignment, and a single subject’s brain may be better than making a mean template from numerous subjects, because it has sharp high-contrast edges, which helps to improve the registration of boundaries {Chiang, 2007 #1564}. Next, a minimal deformation target (MDT), or group mean template, was constructed. This has been advocated in prior studies to reduce bias and improve statistical power (Kochunov et al., 2001; Hua et al., 2008a). To create the MDT, each subject’s 9-parameter registered image is then nonlinearly registered to an average brain template or atlas comprised of 102 individuals from combined developmental imaging studies conducted by the Developmental Cognitive Neuroimaging Group at UCLA, including all typically developing children and adolescents, all subjects with prenatal exposure to drugs of abuse, and all 61 subjects in the present study. We chose to use the larger sample of all subjects in our database, rather than a study-specific MDT, as it may be more representative of the various neurodevelopmental studies in our group, and in the larger developmental neuroimaging community. Following the steps in (Hua et al, 2008), the MDT was constructed by matching one 3D image to the other using our MI-based inverse-consistent algorithm, followed by applying the inverse of the mean displacement field (i.e., inverse geometric centering of the corresponding displacement fields, (Kochunov et al., 2002)) from all subjects to the average 102 average-brain template.

Individual nonlinear registration to MDT

In order to localize differences in brain structure, each individual image is then compared to this MDT by further nonlinearly deforming the anatomy of each individual image to match that of this MDT. Lastly, the Jacobian operator was applied to the deformation fields, whose voxelwise determinants (the Jacobian maps) thus represent the local expansion and contraction factors required to deform the individual’s anatomy to the MDT. These maps represent localized percent-change with positive percentages representing tissue expansion and negative percentages showing tissue contraction. These Jacobian volume deformation maps can then be used to evaluate group differences in voxel-based whole-brain analyses.

Whole-brain Statistical Analysis

We used the computational statistics libraries provided by the Statistics Online Computational Resource (www.SOCR.ucla.edu) to conduct voxel-wise multiple regression analyses throughout the entire brain, where group membership was used to predict Jacobian values at each voxel (Che et al., 2009). These analyses yielded statistical P-value maps (p < 0.05) reflecting the extent to which local volume differed across the 3 groups. Permutation testing (Bullmore et al., 1999) was used to correct for multiple comparisons. In these analyses, subjects were randomly assigned to groups (of the same sample sizes as the original groupings) and statistical maps for 1000 permutations were generated using the same multiple regression analyses used in the original groupings. The percent of significant (p < 0.05) voxels within the brain in the 1000 random groupings was compared to the percent of significant voxels in the real group analyses. Corrected p-values for the observed group difference were created by counting the number of random permutations whose suprathreshold voxels within the entire brain met or exceeded that in the original groupings. This approach has been used in multiple studies (e.g. (Gogtay et al., 2008; Leow et al., 2009)).

In order to investigate how the three groups contributed to the significant group analyses described above, follow-up tests were conducted using three two-sample t-tests (CON vs. ALC, CON vs. MAA, and ALC vs. MAA) using a mask of all significant voxels in the three-group multiple regression analyses. Permutation analyses were conducted to estimate corrected P-values for each contrast.

While there were no statistically significant differences in the gender distribution across the 3 groups, the proportion of boys was higher in the CON than the MAA or ALC groups. Thus, separate statistical analyses were conducted using gender as a nuisance covariate. Further, we ran separate analyses excluding the 3 subjects in the MAA group who did not have concomitant alcohol exposure to ensure that children with only MA exposure were not biasing the results.

Correlations with Full Scale Intelligence Quotient

We extracted percent volume change estimates from each individual’s Jacobian map at each of 14 locations throughout the brain. All regions selected statistically significantly differentiated groups in whole-brain three-group TBM analyses, and included left and right thalamus, left and right caudate nucleus, left and right supracalcarine cortex (posterior cingulate region), right posterior inferior frontal gyrus, right superior parietal lobe, left and right temporal fusiform cortex, left occipital fusiform cortex, right parahippocampal gyrus, and left periventricular callosal white matter (see spatial coordinates in ICBM space in Table 1). Correlations between brain Jacobian volume measures for each of the 14 anatomical points and each subject’s FSIQ score were conducted, and multiple regression analyses were used to examine group-by-FSIQ interactions in predicting Jacobian measures.

Table 1
Jacobian Determinant Regions of interest: anatomical name, ICBM-305 coordinates, hemisphere, location of graph in Figure 1, Jacobian determinant correlation with FSIQ, and Factor Loadings from the discriminant function analyses.

Factor analysis

The same anatomical regions of interest described for the FSIQ-anatomical correlations as well as FSIQ were entered into a classical discriminant function analysis (DFA, Systat Version 12) and used to model the best linear equation among these variables to predict group membership. Factor loadings above 0.3 were considered important for interpretation purposes, and results with and without jackknifed classification are reported. Similar methods have been used in other investigations of the utility of linear combinations of variables in diagnosis of disease states (Ratei et al., 2007)



Demographic descriptors and behavioral performance on FSIQ measures are reported in Table 2. Groups did not differ from each other in age, gender distribution, handedness, or socioeconomic status as estimated from family annual income. The groups differed in FSIQ [F (2,55) = 9.08, p<0.001], with the CON group scoring significantly higher than the ALC and MAA groups, but the MAA and ALC groups did not significantly differ from each other.

Table 2
Demographics for each group (mean and standard deviation).

Three-Group Analyses

Figure 1 maps regions of the brain that were significantly (p < 0.05, uncorrected) different between the 3 groups determined using multiple regression analysis. Numerous brain regions are affected in individuals with prenatal drug exposure including the striatum (as predicted), thalamus, anterior and posterior cingulate, medial and inferior temporal, dorsal and ventral frontal and parietal cortices (as predicted), and cerebellar and posterior callosal white matter. These effects were predominantly bilateral in all cortical and subcortical regions, with 2 exceptions; left but not right hemisphere occipital cortex and right but not left hemisphere frontal pole distinguished groups. Scatterplots of Jacobian volume measurements for each subject from 12 of the 14 regions of interest (regions that significantly differentiated groups in the statistical maps) are shown in Figure 1. Results from whole-brain permutation analyses suggest that the number of significant voxels in the three-group analyses ( at the p = 0.05 level) were not likely observed by chance (Permutation P-value < 0.001).

Figure 1
Statistical F-maps displaying group differences in local brain volume between unexposed controls (CON), alcohol-exposed (ALC), and methamphetamine-exposed (MAA) from the 3-group analyses. Shown are surface renderings with box plots at various points over ...

Two separate multiple regression analyses were conducted, one using group and gender (as a nuisance covariate) to predict regional volumes, and another excluding the 3 subjects in the MAA group who did not have concomitant alcohol exposure. The results did not substantially differ from the 3-group analyses described above.

Follow-up Analyses

Figures 2 and and33 show average volume differences as a percentage of volume reduction in the MAA group relative to controls (Figure 2), and the ALC group relative to controls (Figure 3). The regional pattern of volume increases (yellow orange-red) and volume decreases (shades of blue and purple) in both groups relative to controls is similar in the regions that are increased (or decreased) in volume tended to be increased (or decreased) in both exposed groups relative to controls, though to lesser or greater intensities and spatial extents. Maximum volume reductions are as high as 32% in the ALC group, relative to controls. Volume increases were as high as 27% in the MAA group relative to controls, and as high as 23% in the ALC group relative to controls. The 3 pair-wise comparisons, CON vs. ALC, CON vs. MAA, and MAA vs. ALC are simultaneously mapped in Figure 4, showing regions of statistically significant volume differences between groups in regions where the 3-group analyses were statistically significant (from Figure 1). In this figure, all regions in color were statistically significant in one or more of the two-group statistical analyses for volume increases and volume decreases. Subcortically, thalamic and striatal structures are reduced in volume bilaterally in both groups relative to controls as are left parieto-occipital and right anterior prefrontal cortices. On average, volume reductions in exposed children in these regions are on the order of 10 to 15 percent (shades of blue, and purple in Figure 4), but local variation in these percentages are observed (as shown in the scatterplots in Figure 1). Limbic cortices of the anterior and posterior cingulate, ventral and medial temporal lobes, and bilateral perisylvian cortices are increased in volume bilaterally in both groups relative to controls, an average of approximately 10 to 15 percent averaged across all voxels (orange, yellow and red in Figure 4). It should be noted that the contrast comparing ALC to CON groups was less powerful than the contrast comparing the MAA to CON groups, as there were fewer subjects in the ALC than the MAA group. Thus, the apparent larger extent of significant positive and negative effects in the MAA subjects relative to controls than the ALC subjects relative to controls may be due to the increased statistical power. Whole-brain permutation analyses suggested that the number of significant voxels in both the MAA vs. CON and ALC vs. CON contrasts were not likely observed by chance (Permutation P-value < 0.001 for both contrasts).

Figure 2
MAA percent difference from average CON jacobian values. Maps are color coded according to the color bar where volume increases relative to controls are shown in shades of red, and volume decreases in the MAA relative to CON groups are shown in shades ...
Figure 3
ALC percent difference from average CON jacobian values. Maps are color coded according to the color bar where volume increases relative to controls are shown in shades of red, and volume decreases in the ALC relative to CON groups are shown in shades ...
Figure 4
Follow-up test statistical binarized p-maps at a threshold of P = 0.05 (uncorrected) representing the following contrasts categories between groups where: 1) CON < MAA (red), 2) CON < MAA and CON < ALC (orange), 3) CON < ...

Direct statistical comparison of the 2 exposed groups are alsow shown in Figure 4. To interpret MAA vs. ALC differences, we first examined the direction of group difference in the ALC or MAA vs. control contrast, and then determined which exposed group was more deviant from the controls (in volume reduction or increase), who were considered “baseline normal” for these purposes (centered on zero in the graphs shown in Figure 4). Volume reductions in the striatum, while observed in both groups, appear to be more severe in the MAA group (i.e., MAA volumes significantly smaller than ALC volumes by approxiametly 10%, shown in dark blue in Figure 4). Volume increases were observed in both exposed groups relative to controls in anterior and posterior cingulate regions, and in left and right perisylvian cortices, and these volume increases are more pronounced in the MAA than ALC group (shown in yellow in Figure 4; MAA volume significantly larger than ALC and CON volumes by as much as approximately 15%). While prominent volume increases were observed in both groups in inferolateral and some medial temporal lobe regions, the exposed groups did not significantly differ from each other in these regions. Results from permutation analyses for the direct comparison between MAA and ALC groups were not significant upon permutation correction for multiple comparisons, perhaps due to the smaller samples in the exposed groups relative to the comparisons between each exposed group versus the unexposed controls.

Classical Discriminant Function Analyses (DFA): Classical DFA of the 14 ROI’s and FSIQ revealed 2 factor scores that discriminated the three groups with overall 72% accuracy (67% accurate for the ALC group, 72% accurate for the CON group, and 76% accuracy for the MAA group) using jackknife procedures (see Figure 5). Factor loadings (canonical discriminant functions standardized) on each of the 14 brain and FSIQ predictors are shown in Table 1. The first factor seems most important in determining whether or not an individual was exposed to drugs prenatally, and the second factor differentiates MAA from ALC groups. Qualitative assessment of standardized discriminant function scores for Factor 1 suggest that volumes of the left supracalcarine and occipital fusiform cortex and FSIQ produced the best linear equation to discriminate exposure groups from controls. All of these measures loaded more heavily on Factor 1 than Factor 2. Loadings for Factor 2, discriminating MAA from ALC groups, relied most heavily on right superior parietal lobe, right medial prefrontal, left caudate and right thalamus. All of these regions loaded with factor scores above 0.3 (or below −0.3), and were much higher on factor 2 than factor 1.

Figure 5
A. Factor analysis using Jacobian Values for each of the 14 ROIs (described in Table 1 and illustrated in Figure 1) and full-scale intelligent quotient to predict group membership. B. Group-by-score interactions in predicting left caudate volume, with ...

Brain Volumes and FSIQ: Exploratory analyses for simple correlations between FSIQ and Jacobian-based volume measurements for each of the 14 ROIs shown in Table 1 revealed significant correlations for several ROIs (uncorrected). Left and right occipital (supracalcarine and fusiform) Jacobian volumes were positively correlated with FSIQ such that individuals with lower volumes had lower IQ scores (left r = 0.275, p = 0.036, right r = 0.334, p = 0.018). Volume reductions in the right thalamus were also associated with reductions in FSIQ (r = 0.296, p = 0.024). These findings make sense in light of volume decreases observed in these regions in both exposure groups relative to controls, and suggest that the greater the volume reduction, the lower the intellectual ability. Left inferior temporal fusiform regions were significantly negatively correlated with FSIQ such that individuals with higher volumes had lower FSIQ scores (r = −0.322, p = 0.014). These results also make sense in that volume increases in these inferotemporal regions are observed in both exposed groups relative to controls, and suggest that greater volume increases are associated with greater decrements in FSIQ. While the effects for each brain region that correlated with FSIQ were in the predicted direction, it should be noted that none of these effects remained significant when group membership was used as an additional predictor. This suggests that relationships between brain volumes and FSIQ are mediated by group differences in both brain and FSIQ measures. In other words, individuals with prenatal exposure to drugs of abuse are likely to have lower FSIQ and lower volumes in the thalamus and medial occipital regions than individuals without significant prenatal drug exposure histories. Group-by-score interactions for these measures were significant in predicting the left caudate Jacobian-based volume measurement [F = 3.38, p=0.042] (see Figure 5) such that within the MAA group, smaller volumes were associated with lower FSIQ scores. In contrast to the MAA group, for children in the ALC and CON groups, smaller caudate volumes were associated with higher FSIQ scores.


We have long known that the deleterious effects of prenatal exposure to alcohol on brain morphology endure through childhood and adolescence (Archibald et al., 2001; Bookstein et al., 2001; Autti-Ramo et al., 2002; Bookstein et al., 2002; Sowell et al., 2002; Sowell et al., 2007), and that many women who abuse methamphetamine also abuse alcohol, but here we show for the first time that brain morphology is affected in children with prenatal MA exposure above and beyond the effects of alcohol exposure alone. Overall, the regional pattern of brain dysmorphology was similar in the 2 exposure groups relative to the unexposed controls. That is, regions that were reduced in the ALC group also tended to be reduced in the MAA group. Similar brain systems appear to be affected in similar ways whether the prenatal exposure is to alcohol and not MA, or alcohol and MA. Generally, subcortical structures of the striatum and thalamus, as well as more lateral cortical regions of the temporoparietal lobes bilaterally and the frontal pole in the right hemisphere, are reduced in volume in the exposed relative to control groups. The etiology of volume reductions in these regions is not entirely clear from these structural brain imaging investigations. There is evidence that MA neurotoxicity to brain structures high in monoaminergic receptors (i.e., the striatum) results in agenesis of mature axons and axon terminals (McCann and Ricaurte, 2004), but the impact of MA on developing neural systems likely has a much more pronounced effect. This is because monoaminergic circuits project widely in the brain, and develop early in ontogeny, increasing the likelihood that non-aminergic neural networks could also be impacted (reviewed in (Frost and Cadet, 2000)). Similar mechanisms could be implicated in prenatal alcohol exposure, though different brain neurotransmitter circuits may be targeted (reviewed in (Derauf et al., 2009)). This could help explain the volume reductions in cortical regions without high densities of monoaminergic receptors, and targeting of different neurotransmitter circuits by the 2 different drugs may explain why some brain regions are more or less reduced in volume in the 2 exposure groups.

Striatal volume reductions in children with prenatal MA exposure observed here are consistent with previous findings (Chang et al., 2004), and consistent with our predictions based on the known neurotoxicitiy of MA to dopamine rich brain areas (Volkow et al., 2001), but volume reductions of striatal structures have also been observed in children with heavy prenatal alcohol exposure (Archibald et al., 2001). Striatal volume reductions observed in this sample were more severe in the MAA than ALC groups, suggesting that prenatal MA exposure, or the combination of MA and alcohol exposure has a larger impact on the striatum than alcohol alone. Group by FSIQ interactions were significant in the left caudate, where smaller volume was associated with lower FSIQ scores in the MAA group, and smaller volume was associated with higher FSIQ scores in the CON group. These findings, while preliminary, suggest that the damage to striatal structures resultant from prenatal MA exposure, or to the combination of MA and alcohol exposure, and potential subsequent disturbances in maturational change within those structures is related to lower general intellectual functioning long after the perinatal insult(s) to the brain. This effect appears to be somewhat specific to the striatum as group-by-score interactions were not significant in any other region evaluated, which makes sense given the prevalence of dopamine receptors their vulnerability to MA toxicity. It should be noted, however, that these analyses were exploratory, and should be interpreted with caution until replicated in independent samples.

Volume reductions in left occipital brain regions, right occipital fusiform cortex, and bilateral dorsal parietal cortex were also more severe in the MAA than ALC subjects, and while unexpected, may be consistent with findings of decreased regional cerebral blood flow in left occipital cortices in adult MA abusers (Chang et al., 2002). Local volume increases were observed in both exposure groups relative to the unexposed controls, particularly in the limbic cortices of the anterior and posterior cingulate and inferior and medial temporal and bilateral perisylvian cortices. We have previously noted cortical thickness increases in lateral and medial temporal regions in independent samples of children with heavy prenatal alcohol exposure (Sowell et al., 2007), and others have noted relative volume sparing in medial temporal lobe structures in a similar population (Archibald et al., 2001). This is the first time volume increases in cingulate cortices in children with prenatal MA exposure have been reported, and in a direct comparison of exposed groups, cingulate volume increases were more pronounced in the MAA than the ALC group (i.e., MAA > ALC). The anterior cingulate cortex is part of the attentional network shown to be deficient in children with prenatal MA exposure (Chang et al., 2004), and is associated with monitoring of control, conflict resolution, decision making, and generally connects sensory inputs with executive brain centers in generating motor outputs (reviewed in (Assadi et al., 2009)). The anterior cingulate is also strongly interconnected with dopamine rich striatal, medial temporal, and thalmic structures (reviewed in (Assadi et al., 2009)), all shown to be dysmorphic in children with prenatal drug exposure. Dopamine receptors are most highly expressed in the striatum, but there is evidence that extrastriatal dopamine receptors exist in anterior cingulate and lateral temporal lobe structures (Okubo et al., 1999). While we observe volume increases in extrastriatal limbic cortices in those with prenatal MA and/or alcohol exposure, a report in adult MA abusers show volume deficits in cingulate and medial temporal lobe structures (Thompson et al., 2004). Volume increases in cingulate and other interconnected limbic cortices may be compensatory for dysfunction in striatal and thalamic structures which are reduced in volume in children with MA and/or alcohol exposure. Alternatively, it is possible that volume increases are dysfunctional, resulting from lack of synaptic pruning and/or decreased myelination, cellular changes which are known to continue during childhood and adolescence, and are thought to result in more efficient cognitive processing (reviewed in (Sowell et al., 2003)). While correlations here between FSIQ and local volume in inferior temporal limbic regions and thalamus were significant when considered for all three groups combined, these correlations were largely mediated by group membership. Further study will be required to determine the cognitive significance of brain dysmorphology in children with prenatal exposure to drugs of abuse.

The difference in brain dysmorphology between the MAA and ALC groups highlights that structural brain imaging may yield more information about the integrity of brain networks than evaluations at the behavioral level alone. This statement is further supported by the results from discriminant function analyses presented here.. Whether or not factor scores from the current sample could be used to predict new subjects has yet to be tested, but the estimate from jackknifed analyses of an overall 72% accurate classification is promising. Information from structural brain imaging may eventually be used diagnostically for children with cognitive or behavioral abnormalities but without well-documented exposure histories. That different brain systems are more or less affected in these children with different drug exposure histories suggest that different interventions may be appropriate for cognitive and behavioral deficits depending on which brain systems are more affected.

There are limitations to this study, which necessitate cautious interpretation of results. As with the vast majority of studies of prenatal exposure to drugs of abuse, quantities and frequencies of drug exposure are difficult to accurately recall years after the drug use, and may be compounded by the stigma of admitting to drug use during pregnancy. When women receive methamphetamine from their partners, they may not actually know the dosage they ingest. Potential underreporting by biological mothers is of sufficient concern that reports from adoptive mothers (based on observation of biological mother’s behavior or from social services reports) may have similar levels of validity to that of biological mothers. The majority of our MA-exposed participants also had concomitant alcohol exposure, and it is not clear that quantity and frequency of exposure were comparable between our ALC and MAA groups. The possibility that morphological differences observed in the MAA group is related to an interaction of methamphetamine with alcohol rather than to methamphetamine alone cannot be ruled out. Nonetheless, studying children whose exposure histories are representative of the communities from which they were drawn may be more ecologically valid than trying to find groups of children with “pure” exposures to only one drug.

Also of considerable concern is the concomitant nicotine exposure which is likely more prevalent in the exposed than non-exposed populations. Nicotine exposure is of particular concern because in animal models, it has been shown to disrupt the timing of cell replication and differentiation, and such effects are longer lasting than effects from fetal cocaine exposure (Slotkin, 1998). In a recent survey of 191 prospectively identified MA exposed subjects, approximately 80% were also exposed to nicotine {Della Grotta, 2009 #1558}, and another recently published survey article, 55% of women who reported drinking during pregnancy also admitted smoking {Aliyu, 2009 #1559}. Thus, rates of coexposure to nicotine are high in general in those who abuse MA and/or alcohol, and similar rates are likely present in our samples which would suggest differences between the MAA and ALC groups are not likely due to differences in nicotine exposure. Finally, while mono-drug exposures are rare in humans, there is a vast animal literature suggesting damage to the fetus occurs when alcohol {Sulik, 2005 #867} or MA {Melo, 2008 #1057} are administered alone. Unfortunately, unavailability of nicotine-exposure information from a significant proportion of our exposed subjects makes it difficult to evaluate if the rate of nicotine exposure in the MAA and ALC groups differ. Finally, our sample sizes in the exposed groups were relatively small, but our sample sizes were sufficient to detect the group differences in brain morphology reported here. It remains possible, however that insufficient power precluded us from detecting subthreshold differences between the ALC and MAA groups.


This work was supported by National Institute of Drug Abuse Grants R21 DA15878 and R01 DA017831, National Institute of Alcoholism and Alcohol Abuse grant U01 AA017122, and the March of Dimes (5FY03-12, and 6-FY2008) awarded to ERS. Additional support was provided by the National Center on Research Resources, General Clinical Research Center (3 M01 RR00425) awarded to LMS and National Institutes of Health through the NIH Roadmap for Medical Research, Grant U54RR021813 entitled Center for Computational Biology (CCB; http://nihroadmap.nih.gov/bioinformatics).


Fetal Alchohol Syndrome
Partial Fetal Alcohol Syndrome
Full Scale Intelligence Quotient


  • Archibald SL, Fennema-Notestine C, Gamst A, Riley EP, Mattson SN, Jernigan TL. Brain dysmorphology in individuals with severe prenatal alcohol exposure. Dev Med Child Neurol. 2001;43:148–154. [PubMed]
  • Assadi SM, Yucel M, Pantelis C. Dopamine modulates neural networks involved in effort-based decision-making. Neurosci Biobehav Rev. 2009;33:383–393. [PubMed]
  • Autti-Ramo I, Autti T, Korkman M, Kettunen S, Salonen O, Valanne L. MRI findings in children with school problems who had been exposed prenatally to alcohol. Dev Med Child Neurol. 2002;44:98–106. [PubMed]
  • Berman S, O’Neill J, Fears S, Bartzokis G, London ED. Abuse of amphetamines and structural abnormalities in the brain. Ann N Y Acad Sci. 2008;1141:195–220. [PMC free article] [PubMed]
  • Bookstein FL, Sampson PD, Streissguth AP, Connor PD. Geometric morphometrics of corpus callosum and subcortical structures in the fetal-alcohol-affected brain. Teratology. 2001;64:4–32. [PubMed]
  • Bookstein FL, Streissguth AP, Sampson PD, Connor PD, Barr HM. Corpus callosum shape and neuropsychological deficits in adult males with heavy fetal alcohol exposure. Neuroimage. 2002;15:233–251. [PubMed]
  • Brun CC, Nicolson R, Lepore N, Chou YY, Vidal CN, Devito TJ, Drost DJ, Williamson PC, Rajakumar N, Toga AW, Thompson PM. Mapping brain abnormalities in boys with autism. Hum Brain Mapp 2009 [PMC free article] [PubMed]
  • Bullmore ET, Suckling J, Overmeyer S, Rabe-Hesketh S, Taylor E, Brammer MJ. Global, voxel, and cluster tests, by theory and permutation, for a difference between two groups of structural MR images of the brain. IEEE Trans Med Imaging. 1999;18:32–42. [PubMed]
  • Chang L, Alicata D, Ernst T, Volkow N. Structural and metabolic brain changes in the striatum associated with methamphetamine abuse. Addiction. 2007;102(Suppl 1):16–32. [PubMed]
  • Chang L, Ernst T, Speck O, Patel H, DeSilva M, Leonido-Yee M, Miller EN. Perfusion MRI and computerized cognitive test abnormalities in abstinent methamphetamine users. Psychiatry Res. 2002;114:65–79. [PubMed]
  • Chang L, Smith LM, LoPresti C, Yonekura ML, Kuo J, Walot I, Ernst T. Smaller subcortical volumes and cognitive deficits in children with prenatal methamphetamine exposure. Psychiatry Res. 2004;132:95–106. [PubMed]
  • Chang L, Cloak C, Jiang CS, Farnham S, Tokeshi B, Buchthal S, Hedemark B, Smith LM, Ernst T. Altered neurometabolites and motor integration in children exposed to methamphetamine in utero. Neuroimage 2009 [PMC free article] [PubMed]
  • Che A, Cui J, Dinov ID. SOCR Analyses: Implementation and Demonstration of a New Graphical Statistics Educational Toolkit. Journal of Statistical Software. 2009;30:1–19. [PMC free article] [PubMed]
  • Chiang MC, Reiss AL, Lee AD, Bellugi U, Galaburda AM, Korenberg JR, Mills DL, Toga AW, Thompson PM. 3D pattern of brain abnormalities in Williams syndrome visualized using tensor-based morphometry. Neuroimage. 2007;36:1096–1109. [PMC free article] [PubMed]
  • Cloak CC, Ernst T, Fujii L, Hedemark B, Chang L. Lower diffusion in white matter of children with prenatal methamphetamine exposure. Neurology. 2009;72:2068–2075. [PMC free article] [PubMed]
  • Colby JB, Smith L, O’Connor MJ, Bookheimer SY, Van Horn JD, Sowell ER. A tract-based diffusion imaging study: the effects of prenatal methamphetamine exposure on white matter microstructure Submitted.
  • Colliver JD, Kroutil LA, Dai L, Gfroerer JC. Misuse of prescription drugs: Data from the 2002, 2003, and 2004 National Surveys on Drug Use and Health. Rockville, MD: Substance Abuse and Mental Health Services Administration, Office of Applied Studies; 2006.
  • Derauf C, Kekatpure M, Neyzi N, Lester B, Kosofsky B. Neuroimaging of children following prenatal drug exposure. Semin Cell Dev Biol. 2009;20:441–454. [PMC free article] [PubMed]
  • Frost DO, Cadet JL. Effects of methamphetamine-induced neurotoxicity on the development of neural circuitry: a hypothesis. Brain Res Brain Res Rev. 2000;34:103–118. [PubMed]
  • Glass LA, Ryan JJ, Bartels JM, Morris J. Estimating WISC-IV indexes: proration versus linear scaling. J Clin Psychol. 2008;64:1175–1180. [PubMed]
  • Gogtay N, Lu A, Leow AD, Klunder AD, Lee AD, Chavez A, Greenstein D, Giedd JN, Toga AW, Rapoport JL, Thompson PM. Three-dimensional brain growth abnormalities in childhood-onset schizophrenia visualized by using tensor-based morphometry. Proc Natl Acad Sci U S A. 2008;105:15979–15984. [PMC free article] [PubMed]
  • Hua X, Leow AD, Levitt JG, Caplan R, Thompson PM, Toga AW. Detecting brain growth patterns in normal children using tensor-based morphometry. Hum Brain Mapp. 2009;30:209–219. [PMC free article] [PubMed]
  • Hua X, Leow AD, Parikshak N, Lee S, Chiang MC, Toga AW, Jack CR, Jr, Weiner MW, Thompson PM. Tensor-based morphometry as a neuroimaging biomarker for Alzheimer’s disease: An MRI study of 676 AD, MCI, and normal subjects. Neuroimage 2008a [PMC free article] [PubMed]
  • Hua X, Leow AD, Lee S, Klunder AD, Toga AW, Lepore N, Chou YY, Brun C, Chiang MC, Barysheva M, Jack CR, Jr, Bernstein MA, Britson PJ, Ward CP, Whitwell JL, Borowski B, Fleisher AS, Fox NC, Boyes RG, Barnes J, Harvey D, Kornak J, Schuff N, Boreta L, Alexander GE, Weiner MW, Thompson PM. Alzheimer’s Disease Neuroimaging I. 3D characterization of brain atrophy in Alzheimer’s disease and mild cognitive impairment using tensor-based morphometry. Neuroimage. 2008b;41:19–34. [PMC free article] [PubMed]
  • Kochunov P, Lancaster JL, Thompson P, Woods R, Mazziotta J, Hardies J, Fox P. Regional spatial normalization: toward an optimal target. J Comput Assist Tomogr. 2001;25:805–816. [PubMed]
  • Kochunov P, Lancaster J, Thompson P, Toga AW, Brewer P, Hardies J, Fox P. An optimized individual target brain in the Talairach coordinate system. Neuroimage. 2002;17:922–927. [PubMed]
  • Lebel C, Rasmussen C, Wyper K, Walker L, Andrew G, Yager J, Beaulieu C. Brain Diffusion Abnormalities in Children With Fetal Alcohol Spectrum Disorder. Alcohol Clin Exp Res 2008 [PubMed]
  • Lee AD, Leow AD, Lu A, Reiss AL, Hall S, Chiang MC, Toga AW, Thompson PM. 3D pattern of brain abnormalities in Fragile X syndrome visualized using tensor-based morphometry. Neuroimage. 2007;34:924–938. [PMC free article] [PubMed]
  • Leow A, Huang SC, Geng A, Becker J, Davis S, Toga A, Thompson P. Inverse consistent mapping in 3D deformable image registration: its construction and statistical properties. Inf Process Med Imaging. 2005;19:493–503. [PubMed]
  • Leow AD, Yanovsky I, Chiang MC, Lee AD, Klunder AD, Lu A, Becker JT, Davis SW, Toga AW, Thompson PM. Statistical properties of Jacobian maps and the realization of unbiased large-deformation nonlinear image registration. IEEE Trans Med Imaging. 2007;26:822–832. [PubMed]
  • Leow AD, Yanovsky I, Parikshak N, Hua X, Lee S, Toga AW, Jack CR, Jr, Bernstein MA, Britson PJ, Gunter JL, Ward CP, Borowski B, Shaw LM, Trojanowski JQ, Fleisher AS, Harvey D, Kornak J, Schuff N, Alexander GE, Weiner MW, Thompson PM. Alzheimer’s disease neuroimaging initiative: a one-year follow up study using tensor-based morphometry correlating degenerative rates, biomarkers and cognition. Neuroimage. 2009;45:645–655. [PMC free article] [PubMed]
  • Lu LH, Johnson A, O’Hare ED, Bookheimer SY, Smith LM, O’Connor MJ, Sowell ER. Effects of prenatal methamphetamine exposure on verbal memory revealed with functional magnetic resonance imaging. J Dev Behav Pediatr. 2009;30:185–192. [PMC free article] [PubMed]
  • McCann UD, Ricaurte GA. Amphetamine neurotoxicity: accomplishments and remaining challenges. Neurosci Biobehav Rev. 2004;27:821–826. [PubMed]
  • Mooney SM, Miller MW. Effects of prenatal exposure to ethanol on the expression of bcl-2, bax and caspase 3 in the developing rat cerebral cortex and thalamus. Brain Res. 2001;911:71–81. [PubMed]
  • Okubo Y, Olsson H, Ito H, Lofti M, Suhara T, Halldin C, Farde L. PET mapping of extrastriatal D2-like dopamine receptors in the human brain using an anatomic standardization technique and [11C]FLB 457. Neuroimage. 1999;10:666–674. [PubMed]
  • Ratei R, Karawajew L, Lacombe F, Jagoda K, Del Poeta G, Kraan J, De Santiago M, Kappelmayer J, Bjorklund E, Ludwig WD, Gratama JW, Orfao A. Discriminant function analysis as decision support system for the diagnosis of acute leukemia with a minimal four color screening panel and multiparameter flow cytometry immunophenotyping. Leukemia. 2007;21:1204–1211. [PubMed]
  • Riley EP, McGee CL. Fetal alcohol spectrum disorders: an overview with emphasis on changes in brain and behavior. Exp Biol Med (Maywood) 2005;230:357–365. [PubMed]
  • Shattuck DW, Leahy RM. BrainSuite: an automated cortical surface identification tool. Med Image Anal. 2002;6:129–142. [PubMed]
  • Sled JG, Zijdenbos AP, Evans AC. A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Transactions on Medical Imaging. 1998;17:87–97. [PubMed]
  • Slotkin TA. Fetal nicotine or cocaine exposure: which one is worse? J Pharmacol Exp Ther. 1998;285:931–945. [PubMed]
  • Smith LM, Chang L, Yonekura ML, Grob C, Osborn D, Ernst T. Brain proton magnetic resonance spectroscopy in children exposed to methamphetamine in utero. Neurology. 2001;57:255–260. [PubMed]
  • Smith LM, LaGasse LL, Derauf C, Grant P, Shah R, Arria A, Huestis M, Haning W, Strauss A, Della Grotta S, Liu J, Lester BM. The infant development, environment, and lifestyle study: effects of prenatal methamphetamine exposure, polydrug exposure, and poverty on intrauterine growth. Pediatrics. 2006;118:1149–1156. [PubMed]
  • Sowell ER, Peterson BS, Thompson PM, Welcome SE, Henkenius AL, Toga AW. Mapping Cortical Change Across the Human Life Span. Nature Neuroscience. 2003;6:309–315. [PubMed]
  • Sowell ER, Mattson SN, Kan E, Thompson PM, Riley EP, Toga AW. Abnormal Cortical Thickness and Brain-Behavior Correlation Patterns in Individuals with Heavy Prenatal Alcohol Exposure. Cereb Cortex 2007 [PMC free article] [PubMed]
  • Sowell ER, Thompson PM, Mattson SN, Tessner KD, Jernigan TL, Riley EP, Toga AW. Regional brain shape abnormalities persist into adolescence after heavy prenatal alcohol exposure. Cerebral Cortex. 2002;12:856–865. [PubMed]
  • Thompson PM, Hayashi KM, Simon SL, Geaga JA, Hong MS, Sui Y, Lee JY, Toga AW, Ling W, London ED. Structural abnormalities in the brains of human subjects who use methamphetamine. J Neurosci. 2004;24:6028–6036. [PubMed]
  • Volkow ND, Chang L, Wang GJ, Fowler JS, Leonido-Yee M, Franceschi D, Sedler MJ, Gatley SJ, Hitzemann R, Ding YS, Logan J, Wong C, Miller EN. Association of dopamine transporter reduction with psychomotor impairment in methamphetamine abusers. Am J Psychiatry. 2001;158:377–382. [PubMed]
  • Wechsler D. The Wechsler Intelligence Scale for Children. 4. San Antonio, TX: The Psychological Corporation; 2003.
  • Wells WM, 3rd, Viola P, Atsumi H, Nakajima S, Kikinis R. Multi-modal volume registration by maximization of mutual information. Med Image Anal. 1996;1:35–51. [PubMed]
  • Yanovsky I, Leow A, Lee S, Osher SJ, Thompson PM. Comparing registration methods for mapping brain change using tensor-based morphometry. Medical Image Analysis In Press. [PMC free article] [PubMed]
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...