An MRI and Head Circumference Study of Brain Size in Autism: Birth through Age Two Years

Author: Dr Heather Cody Hazlett - University of North Carolina, USAHeather Cody Hazlett PhD is an Assistant Professor in the Department of Psychiatry at the University of North Carolina, Chapel Hill. She is a psychologist with a background in paediatric neuropsychology.  Her interests include using brain- imaging techniques, such as MRI, to study brain development in developmental disorders such autism and Fragile X. 

Abstract: Heather Cody Hazlett, Ph.D.*, Michele Poe, Ph.D., Guido Gerig, Ph.D.,Rachel Gimpel Smith, B.A., James Provenzale, M.D. [1], Allison Ross, M.D. [2], John Gilmore, M.D., and Joseph Piven, M.D

UNC Neurodevelopmental Disorders Research Center he University of North Carolina at Chapel Hill

[1] Department of Radiology, [2] Department of Anesthesiology, Duke University Medical Center

* To whom correspondence should be addressed:     Heather Cody Hazlett, Ph.D.     UNC Neurodevelopmental Disorders Research Ctr     University of North Carolina, CB#3367     Chapel Hill, NC 27599-3367     heather_cody@med.unc.edu

Research supported by NIH grants MH61696 (J Piven), HD03110 (J Piven), and EB002779 (G Gerig)Word length:  5459

Background:  While the neuroanatomical basis of autism is not yet known, evidence suggests that brain enlargement may be characteristic of this disorder. Inferences about the timing of brain enlargement have recently come from studies of head circumference.Objective:  To examine brain volume and head circumference in autistic individuals compared to controls.Design:  A cross-sectional study of brain volume was conducted at the first time-point in an ongoing longitudinal MRI study of brain development in autism. Retrospective longitudinal head circumference measurements were gathered from medical records on a larger sample of autistic individuals and local controls.Setting:  Clinical research center.Participants:  The MRI study included 51 autistic and 25 control children between 18-35 months of age (the latter included both developmental delayed and typically-developing children). Retrospective, longitudinal head circumference data were examined from birth to age 3 years in 113 autistic and 189 local control children.Main Outcome Measures:  Cerebral cortical (including cortical lobes) and cerebellar gray and white matter MRI brain volumes, and retrospective head circumference from medical records were studied.Results:  Significant enlargement was detected in cerebral cortical but not cerebellar volumes in autistic individuals. Enlargement was present in both white and gray matter and was generalised throughout the cerebral cortex. Head circumference appears normal at birth with a significantly increased rate of head circumference growth appearing to begin around 12 months of age.Conclusions:  Generalized enlargement of gray and white matter cerebral volumes, but not cerebellar volumes, are present at two years of age in autism.  Indirect evidence suggests that

this increased rate of brain growth in autism may have its onset post-natally in the latter part of the first year of life.

Full Paper:

Introduction

Autism is a complex neurodevelopmental disorder defined by the presence of social deficits, abnormalities in communication, the presence of stereotyped, repetitive behaviors, and a characteristic course (1).  While the neuroanatomical basis of this condition is not yet known numerous lines of evidence suggest that abnormalities in brain volume that may be characteristic of autism.  Head circumference studies in children and adults with autism have consistently identified a subset of approximately 20% with macrocephaly (i.e., greater than the 98th percentile for head circumference) (2-4) and two post mortem studies have shown a high proportion of autistic individuals with increased brain weight (5, 6).

We first reported increased brain size on MRI in two independent samples of adolescents and adults with autism, in comparison to age, gender and IQ comparable controls (7, 8).  More recently Courchesne et al (9) examined a cross-sectional sample of 60 autistic males (IQ 36-122) and 52 typically-developing male controls (IQ > 85), ages 2 to 16 years.  Post-hoc analysis revealed enlargement of gray and white cerebral volume in autistic individuals 2 through 4 years of age, but were actually decreased in autistic individuals 5 to 16 years of age.  White matter cerebellar volumes were increased in the 2-4 year olds but significantly decreased in adolescence, while cerebellar gray matter volumes were slightly (but not significantly) increased in the 2-4 year old range and decreased in adolescence.  Sparks et al. (10), reported increased brain volume on MRI in 3-4 year olds with autism compared with both a typically-developing and a developmentally-delayed, non-autistic comparison group, but gray and white matter volumes were not examined.   Cerebellar volumes were increased in the subjects with autism compared to the typically developing controls.  

Aylward et al. (11) examined a large sample of high functioning (IQ > 80) autistic individuals and typically-developing controls age 8-46 years.  No differences were found in overall total brain volume.  However, increased brain volume was observed in a subset of autistic subjects under 12 years of age.  No differences were detected in the older age group.  A recent study by Herbert et al (12) of 17 high-functioning autistic males age 7-11 years of age revealed increased white matter volume in autistic individuals compared to 15 typically developing controls, after controlling for total brain size.  Finally, Lotspeich et al (13) has recently reported increased gray matter, but not white matter, volume in the cerebral cortex of 31 high and low functioning autistic individuals, 8 to 18 years of age.  

While the finding of brain enlargement on MRI has been consistent across all of these studies and consistent with head circumference and post-mortem reports, the distribution (across structures, regions and tissues) (12,13), associated demographic features (e.g., gender, IQ), course and clinical correlates of this phenomenon are not yet clear.  Methodological differences between studies make comparisons difficult.  Perhaps of greatest importance in contributing to the variation across studies, however, is the use of cross-sectional data to make inferences about longitudinal, developmental morphology brain changes.  As demonstrated in the important study by Giedd et al (14), findings from a longitudinal design can provide substantially different results than cross-sectional analyses of the developing brain.

Inferences about the timing of brain enlargement have recently come from retrospective, longitudinal studies of head circumference.  In a retrospective chart review study of autistic individuals with macrocephaly, Lainhart et al (3) found evidence suggesting macrocephaly was not present at birth.  Other studies have found results consistent with this finding (4, 9, 15).   Most recently, Courchesne et al. (16) reported on retrospective, longitudinal head circumference data

from a small sample of autistic individuals (15 autistic individuals examined at four time points and seven subjects with data from birth plus one other time point) versus normative data available from the population and noted that autistic individuals in this study had a decreased head circumference at birth, increased head circumference at 6-14 months of age and a subsequent decreasing rate of head circumference growth with increasing age.  Previous studies, with the exception of Bailey et al. (2), have also compared head circumference in autistic individuals to available population norms. 

In the present study, we sought to address some of these issues by conducting a longitudinal study of MRI gray and white matter brain volumes in a large sample of two year old autistic children and a comparison group which includes both children with typical development and developmental delay (without evidence of autism). These children are part of a longitudinal study and will be examined again at age 4 years.  We report here the results of our 'time one' cross-sectional study of subjects between ages 18-35 months. In addition, we examined a longitudinal, retrospective head circumference dataset on a large sample of autistic individuals, diagnosed with semi-structured measures, and local controls comparable on exclusion criteria, gender, ethnicity and body mass, to provide insight into the timing of brain size changes in this disorder.

Methods

Sample  

Subjects included 51 children with autism and 25 comparison cases, 18-35 months of age.  In the comparison group, there were 11 children with developmental delay (DD) without evidence of a pervasive developmental disorder and 14 children with typical development (TYP).  The children with developmental delay were included to enrich the control sample with subjects that were comparable to the autism subjects in cognitive development.  Children with autism were primarily referred from nine specialty clinics for pervasive developmental disorders in North Carolina (TEACCH, Treatment and Education of Autistic and related Communication Handicapped Children); DD children were referred from selected regional state Children's Developmental Services Agencies (CDSA).  TYP children were recruited from community advertisements.  Subjects with autism were referred after receiving a clinical diagnosis of an autism spectrum disorder.  Subjects with DD were referred only if they had no known identifiable cause for their delay (e.g., prematurity, genetic or neurological disorder) and had no indication of a pervasive developmental disorder.  All subjects were excluded for evidence of a medical condition thought to be associated with autism (17), including Fragile X Syndrome (FraX), Tuberous Sclerosis (TS), gross CNS injury (e.g., cerebral palsy, significant complications or perinatal/postnatal trauma, drug exposure), seizures, and significant motor or sensory impairments. Study approval was acquired from both the UNC and Duke Institutional Review Boards and written informed consent was obtained by getting parental (or custodial guardian) consent for each subject.  All subjects in the MRI study were included in the head circumference study.  An additional 62 subjects with autism and 164 TYP children were added to improve the power needed for the statistical model.  The total sample therefore included 113 subjects with autism and 189 controls (11 DD subjects, and 178 TYP subjects).  The additional autistic subjects were obtained from other studies conducted by our group (18, 19) using the same diagnostic and exclusionary criteria employed in the current study.  TYP subjects were selected from the children born at UNC Hospitals between 2000 and 2002, who also had medical records available from the UNC outpatient clinic through 2004 (2-4 years of age).   Medical records from these controls were screened and subjects were excluded for evidence of prematurity (gestational age <37 months), genetic or neurological disorder (including FraX, gross CNS injury, cerebral palsy, significant obstetric complications or perinatal/postnatal trauma, drug exposure, seizures or significant motor or sensory impairments).  Birth and well baby visits were reviewed and for each available visit head circumference, height, and weight were recorded.  Additional descriptive data included birth date, gender, race, ethnicity, home zip code, parental education level and occupation, and

number of siblings. IRB permission was obtained to perform this hospital records review.  Additional reference data, not employed in the case-control analyses in this study, were obtained from the Centers for Disease Control (CDC) head circumference dataset (20). 

Clinical Assessment

Subjects with autism were enrolled between 18 and 35 months of age.  Medical records and developmental history were reviewed.  Diagnosis was confirmed using the Autism Diagnostic Interview-Revised (21) and the Autism Diagnostic Observation Schedule-G (22).  Subjects were included if they met ADI-R algorithm criteria for autism, and obtained ADOS-G scores consistent with autism.  All cases met DSM-IV criteria (1) for autistic disorder. At time 1 our 18-35 month old sample falls below the recommended age range for employment of these measures in the diagnosis of autism.  In the design of our study, diagnosis will be reassessed at the more conventional time for standardised assessment of autism with these instruments, age 4 years (42-59 months) and it is possible that some subjects may no longer meet full criteria for autistic disorder at that time. 

All subjects were given a battery of measures, including the Mullen Scales of Early Learning (23), the Vineland Adaptive Behaviour Scales (24), Preschool Language Scale 4th edition (25), behavioural rating scales, and a standardized neurodevelopmental examination to exclude subjects with any notable dysmorphology, evidence of neurocutaneous abnormalities, or other significant neurological abnormalities (e.g., lateralized deficits).  All autistic and DD subjects received testing (cytogenetics or molecular) for Fragile X Syndrome.  DD and TYP children were screened for autism with the Childhood Autism Rating Scale (26) and excluded if they reached the cutoff for autism (> 30 total score). Medical records were also reviewed for any possible evidence of autism or PDD-NOS and subjects were excluded from this group for any suggestion of these disorders.  Table 1 displays the subject characteristics (e.g., gender, age, cognitive ability, and adaptive functioning) of the final study population.

 

MRI Acquisition

All subjects were scanned at the Duke-UNC Brain Imaging and Analysis Center (BIAC), on a 1.5 Tesla GE Signa MRI scanner.  Image acquisition was designed to maximise gray/white tissue contrast for the 18-35 month old child and included:  (1) a coronal T1 IR Prepared:  T1 300 msec, TR 12 msec, TE 5 msec, 20o flip angle, at 1.5 mm thickness with 1 NEX, 20 cm FOV; and 256 x 192 matrix;  (2) a coronal PD/T2 2D dual FSE, TR 7200 msec, TE 17/75 msec, at 3.0 mm thickness with 1 NEX, 20 cm FOV, and 256 x 160 matrix.  A series of localizer scans and a set of phantoms was used to standardize assessments over time and individuals.

Subjects with autism and DD were scanned using moderate sedation (combination of pentobarbital and fentanyl as per hospital Sedation Protocol) administered by a sedation nurse and under the supervision of a pediatric anesthesiologist.  Physiological monitoring was conducted throughout the scan and recovery.  TYP subjects were scanned without sedation, in the evening, while sleeping.  All scans were reviewed by a paediatric neuroradiologist and screened for significant abnormalities (e.g., malformations, lesions, etc.).  Seven scans were excluded from the total brain volume measures for motion artifact or acquisition problems.  An additional scan was excluded from the regional and lobe analyses for an error in image acquisition.

Image Processing

Initial image processing to register and align the T1 and PD/T2 scans into a standardised plane was conducted with BRAINS2 developed at the University of Iowa (27-33).  All scans were registered along an AC-PC (anterior-posterior commissure) axis.   For a small number of scans with suboptimal PD/T2 scan quality, a fit-interleave correction using BRAINS2 was also performed.   The co-registered and aligned images were then processed for tissue segmentation using the Expectation Maximization Segmentation (EMS) software [originally developed at the Catholic University of Lueven (34, 35) and adapted by our lab (36)].  Reliability and validity of the EMS software has been rigorously examined by the developers and within our lab (34, 35, 37, 38).  Our initial attempts to apply the existing, adult based, EMS template brain atlas provided unsatisfactory results.  Therefore, a new pediatric template atlas was created by our lab using MRI brain scans of 14 children (comprised of 9 autism, 2 DD, and 3 NC cases randomly selected) that were first tissue classified using BRAINS2, which provides semi-automated tissue classification procedures, and then averaged to create a probabilistic spatial prior template. This resulted in an 'averaged' probabilistic brain atlas that was aligned to each subject brain using a linear, affine transformation in a fully automated procedure.  After bias estimation, inhomogeneity correction, and non-brain stripping procedures were conducted, subject scans were processed with EMS to produce gray, white, and CSF tissue segmented images for each subject.   Total brain volume (TBV) measures included total gray and white matter and all CSF.  Total tissue volume (TTV) included all gray and white matter in the cerebrum and cerebellum.

Regional lobe measurements were obtained using a manually parcellated template (atlas) MRI of a two year old brain developed by our group (39). This scan was randomly selected from our database based on having good image quality. Anatomical landmarks were chosen based on standard neuroanatomy references (40-45), consultation with pediatric neuroradiology experts, and any existing protocols from the literature.  Delineated regions included the frontal, temporal, parietal, and occipital lobes, cerebellum, corpus callosum, and interhemispheric fissure (see Table 2 for description of boundaries).  A 'subcortical area' was produced to exclude basal ganglia, thalamus, deep white matter, and brainstem from all other regions.  This area was not used to perform any regional comparisons, as other image processing methods are more appropriate and successful in subdividing this area.   The insula and cingulate gyrus were also defined, but for the purposes of these analyses the insula was included in the cerebrum measure and cingulate with the frontal/parietal lobes. 

The template brain was then mapped onto the T1 images from our dataset [after being adjusted for intensity differences and affine registered using the Rview program (46)] to obtain label maps for all these regions using a fluid high-dimensional deformation algorithm (47-49) which is completely automated.  The results of our construct and criterion validity studies show that the output from this protocol is both anatomically valid and closely approximates a manual parcellation.  For example, the cerebrum, cerebellum, and frontal lobe regions produced highly consistent volumes with warping alone (intraclass correlation coefficients of overlap between warped volume and manual trace volume ranging from .97 to .99).  Our review of the results revealed difficulty separating the borders between the parietal and occipital lobes and so these two lobes were combined for our analyses of this region.  Label maps were then combined with

the EMS tissue classified images to produce gray/white/CSF volumes for each of these lobe compartments.

 

Statistical Analyses  

A priori hypotheses were tested using mixed models with repeated measures over the regional brain volume domains.  In all models the regional volume was the dependent variable with diagnostic group as the predictor of interest and age and gender as covariates.  Diagnostic group was entered as a 3 level categorical variable (autism, DD, TYP).  Separate parameters were estimated for the DD and the TYP group.  A combined estimate for ‘controls’ was then created using a weighted average of the 2 control groups.

Three sets of analyses were performed examining segmented brain volume and autism diagnosis.  The first set of analyses examined brain volume comparisons of total gray matter, white matter and CSF.  For this model, there were 3 observations per individual (gray, white, and CSF volume) with a categorical variable indicating tissue type. Estimates for TBV volume were created by combining gray, white, and CSF volumes, while estimates for total tissue volume were created combining the estimates for gray and white matter volumes. The second set of analyses examined regional brain volumes for the cerebellum and cerebrum.  In this analysis there were 8 observations per individual. The model included 3 categorical variables indicating region, hemisphere (left/right), and tissue (gray/white).  The third set of analyses examined lobe volumes of the cerebrum.  In this analysis there were 12 observations for each individual that were defined by hemisphere (left/right), tissue (gray/white), and lobe (frontal, temporal, parietal-occipital).

Age, gender, and group were included as predictors in each of the models along with all 2-, 3- and 4-way interactions between lobe, hemisphere, tissue and group, the 2-, 3-, and 4-way interactions between gender and hemisphere, tissue and lobe, and all 2- and 3-way interactions between age, hemisphere, tissue, and lobe.

Laterality indices for gray and white tissue volumes were calculated using the following formula:  [L-R/(L+R) x 100], where left hemisphere laterality is indicated by positive (+) values and right hemisphere laterality is indicated by negative (-) values.  Group, age and gender were regressed on laterality for each lobe and tissue.

The retrospective, longitudinal study of head circumference was performed to compare growth trajectories for head circumference between the groups.  A non-linear mixed model was fit using a exponential growth function [y = b0 + b1(e(b2*age))] with  b0 as the random effect.   All observations were limited to those at 4 years of age and younger.  The mean number of observations per group was 4 for the autism and TYP groups and 7 for the DD group.  Reliability of the retrospective head circumference data was unknown as this was all taken from hospital medical records.  Two models were fit, the first included only race and gender as covariates, the second added Body Mass Index (BMI) as an additional covariate. An estimate of socioeconomic status (maternal education) was intended to be included in the model.  However, the majority of TYP medical records were missing this information and the remaining subset was too small to meaningfully examine this factor.  The inclusion of subject characteristics (e.g., BMI, gender, race) was thought to control for any major group differences from these factors.     

Results

 A description of the sample (N, age, gender, IQ) appears in Table 1.   Group differences were evaluated for age, gender, adaptive functioning (Vineland), and IQ (from the Mullen).  Gender was unequally distributed across groups and was therefore  included as a covariate in all the analyses.  We did not have a large enough group of females to perform separate analyses by group for the brain volumes.  Age differences were also observed (with the TYP group being slightly younger) and was therefore also included as a covariate in the analyses.  The inclusion of the DD and TYP groups were sufficient to control for IQ differences, as IQ was not found to be a significant predictor when considered simultaneously with these groups.

Total Brain Volume

Adjusted means for total brain volume (TBV), total tissue volume (TTV), total gray matter volume (TGM), total white matter volume (TWM), and total cerebral spinal fluid volume (CSF) for the autism, TYP and DD groups are reported in Table 3.  Subjects with autism had significant enlargement in TBV, TTV, TGM, TWM, and marginal evidence suggesting enlargement of CSF (p = .053) compared to controls.  Percent increases in brain volumes in autistic subjects over the

combined group of controls range from 4% to 6%.  Mean differences, standard errors, significance tests and percent differences are reported in Table 4.

Group comparisons were performed for the cerebral cortical and cerebellar volumes between the autistic and control subjects.  No significant differences in the group by side effect (F7,71 = 1.03; p = .42) were detected, therefore left/right differences by region and tissue type are not reported.  Mean and adjusted brain volumes are reported in Table 3 for the group comparisons.  Children with autism had a significant enlargement in total cerebral cortical volume (mean volume increased 5.7% versus controls); however, significant differences were not detected in cerebellar volume (total, gray or white matter) versus controls.  Gray and white matter cerebral volumes were significantly larger in autistic subjects than controls. 

Although we acknowledge the small size of the DD and TYP control subgroups for separate analysis, further exploratory comparisons were made with autistic subjects to gain insight into the specificity of the observed effects for autism rather than generalized developmental delay (or mental retardation), as autistic individuals commonly have co-occurring mental retardation.  Subjects with autism had significant enlargement in TBV, TTV, TGM and TWM compared to the DD subgroup.   In contrast, while the autistic subjects had increased brain volumes compared to the TYP group the difference did not reach the level necessary for statistical significance.  Differences were also not detected between the DD and TYP groups.  Compared with the DD subgroup, autistic subjects had significant enlargement of total cerebral volume as well as cerebral GM and WM volumes.  A graph of cerebral volumes appears in Figure 1.  No significant cerebellar volume differences were found in any of the group comparisons (autism v. DD or TYP).  

Cerebral Cortical Lobe Volume

Mean adjusted GM and WM lobe volumes are reported in Table 5 for the groups.  Autistic subjects had significant enlargement in frontal, temporal and parietal-occipital lobes, for both GM and WM, versus the combined control group.   Percent differences, between autistic and control subjects, in total tissue volumes across the lobes ranged from 3.5% to 9% (see Table 6).  In general, per cent increases in WM volumes were greater than GM volumes.  The largest percent increase was seen in the temporal lobe for both WM and GM. 

Compared to the DD group, the autism group had significantly enlarged left and right temporal and frontal GM volumes.  Parietal-occipital GM was not significantly enlarged but approached significance (p=.06, for both left and right).  The autism group also had significantly enlarged left

and right temporal, frontal, and parietal-occipital WM volumes (see Table 6) compared to the DD group. For the autism and TYP comparison, left and right temporal and parietal-occipital WM were significantly enlarged in the autism group.  Right temporal gray was also enlarged but not significant (p=.06). 

Head Circumference

Growth trajectories of head circumference were examined using longitudinal ratings from birth to 4 years of age.  The average number of head circumference ratings available from medical records, across the 0-4 year old age interval, were as follows:  autism = 4, DD = 7, TYP = 4.  The correlation between head circumference and TBV for the sample was p = .88.   Overall the growth curves for autistic and control subjects, displayed in Figure 2, were significantly 2= -.03; p1=1.65, 0= -1.67, different ( < .001 for all).  No group differences between head circumference ratings were found at birth.  The two curves begin to diverge at around 12 months of age, with the autism group showing an increased rate of head circumference growth, which continues throughout the rest of the age interval.

Body Mass Index (BMI) was calculated on a subset of the sample (autism = 41, DD = 11, TYP = 135) to take differences in body size in to account and address the specificity of the HC effect.  Differences in head circumference between the autistic and TYP groups was unchanged versus the comparisons with the inclusion of BMI, suggesting that the observed effects in HC and MRI brain volumes were not the result of a generalized enlargement in body size.   For comparison to other studies we also examined HC in our subject groups versus the normative data published by the CDC.  In this comparison the autism group appears to have a slightly smaller HC at birth than the CDC group but otherwise no significant differences were observed.

Discussion

In the present study, we find evidence of brain enlargement in a relatively large sample of two year olds with autism compared to a control group which includes both typically developing and developmentally delayed children.  This approximate 5% overall enlargement appears primarily to be the result of increases in both gray and white matter volumes of the cerebral cortex.  In general, we observed brain volume enlargement in both the gray and white matter tissue compartments, although white matter volume increases appeared somewhat more robust than proportionate differences seen in gray matter.  These findings are largely consistent with those of Courchesne et al. (9) who reported gray and white matter enlargement in 2-4 year olds and Sparks et al. (10) who found enlargement in 3-4 year olds with autism.  However, as compared to both of those reports, we failed to find any evidence of enlargement in cerebellar volume, in either total, gray or white matter volumes.  These findings also differ from our own previous report of overall cerebellar volume enlargement in an adolescent and adult sample (7).  Of note, our current sample differs from the Courchesne et al. (9) and Sparks et al. (10) samples where the predominant age of subjects was 3-4 years of age. It is possible that cerebellar enlargement is a later occurring effect that we will detect at follow-up of our current 2 year old cohort at age 4 years.  Alternatively, differences may result from differences in image processing.  In the present study, we use an automated high-dimensional warping method based on an atlas of the two year

old brain, whereas in both the Sparks et al. (10) and Courchesne et al. (9) studies, manual tracings of the cerebellum were conducted.  In our previous study cerebellar volume was measured using an automated system based on the Talaraich coordinates that we believe is less anatomically valid than our current approach.Cerebral cortical enlargement in our sample appeared to be the result of a generalized enlargement in the three cortical regions examined (frontal, temporal and parietal-occipital lobes) although percent enlargement was greatest in temporal lobe.  In a previous report by our group, in a study of adolescents and adults with autism, again using the Talaraich coordinates for automated measurement of lobe volumes, we did not detect significant enlargement in frontal lobes (8).  In a recent re-examination of this earlier adolescent-adult sample, however, using an updated version of the same Talaraich co0ordinates software that allowed us to segment gray and white matter volumes (not previously studied), we detected enlargement in frontal, temporal and parietal lobes but not occipital lobe (50), paralleling findings reported by Carper et al. (51) in 2-3 year olds, using a manual method for lobe parcellation.  Our semi-automated method did not allow us to confidently separate parietal from occipital regions, and we are therefore unable to comment on whether differences we detected are the result of enlargement of the parietal lobe and/or occipital lobe. Given the large inter-individual variation in autism, such differences will need to be examined prospectively in a longitudinal analysis before more definitive conclusions about patterns across lobes and tissues, over time, can be made with confidence.

Our combined control group took into account the fact that autism is often associated with mental retardation, which may itself have an effect on brain development separate from that seen in autism without mental retardation.  While we are hesitant to make too much of exploratory analyses between autistic subjects and our relatively small samples of TYP (N=14) and DD (N=11) control subgroups, differences in these comparisons were observed and warrant comment.    Brain enlargement in autistic subjects in our study was more readily apparent when autistic subjects were compared to the DD group than in comparison to the TYP group.  If, as we believe, mental retardation is important to take into consideration in MRI studies of the brain in autism, then based on our comparison of autistic subjects and our small group of individuals with DD, it can be concluded that the effect of autism is generalized across both gray and white matter compartments.  However, if observations were limited to a comparison with TYP two year olds only, then one would conclude that the autism effect is more specific to white matter, limited to temporal and parietal-occipital regions of the cortex and somewhat less robust.  Of note, no significant relationship between IQ and brain volume was detected in autistic individuals, suggesting that IQ is not important and that observations in this study may be the result of abnormalities or bias in the DD group.  However, the standard deviation in brain volumes of the DD group was actually less than that seen in the TYP group, somewhat going against the idea that the DD group may have been more heterogeneous and less representative.  It is also worth noting that evaluation of IQ at this early age, in our sample, is problematic.  Almost all of the two year olds with autism in our sample showed evidence of low IQ, as children identified with autism at very young ages are more noticeably impaired than higher-functioning autistic individuals, who are often not detected until later ages.  Overall, more definitive interpretations regarding the effect of IQ in MRI studies of autism await studies employing larger comparison groups of both TYP and DD individuals followed forward over time.  In the present study we detected no evidence of cerebral laterality. In our recent re-analysis (50) of an earlier sample (8) we observed left laterality (disproportionately greater in autistic subjects than that which is normally present in controls) in temporal lobe gray matter.  Longitudinal analysis of volume changes over time may reveal that this effect emerges at older ages.  Examination for clinical correlates using global measures (ADI-R algorithm subdomain scores: Social, Communication, Ritualistic Repetitive Behavior, Abnormal Development) aimed more specifically at diagnosis than measurement of specific symptom domains in two year olds did not reveal any obvious brain-behavior correlations.  As noted above, IQ effects were also not observed in this young sample.  Gender effects were not examined as we had too few girls with autism to explore this confidently.  

We examined retrospective, longitudinal data on head circumference in a large sample of autistic individuals, diagnosed with standardized measures. We compared these subjects to a large sample of locally ascertained controls, using exclusion criteria comparable to autistic cases and taking several potential confounding variables into account, including gender, race, and body size. We confirmed the significant enlargement of head circumference in autism and found strong evidence to suggest that the increased trajectory of head circumference growth has its origin during the post-natal period around 12 months of age.  We acknowledge that subtle differences in this estimate can occur given the statistical model employed and that greater differences (i.e., an earlier onset of enlargement) may possibly be detected if the comparison group includes a larger sample of developmentally-delayed, non-autistic individuals.  These findings are not inconsistent with those of Courchesne et al (52) who detected significant enlargement in an age interval between 6 and 14 months.

The findings from this study confirm the presence of generalised cerebral cortical gray and white matter brain volume enlargement at age two in autistic individuals.  Given the strong relationship between head circumference and brain volume (9), the onset of this enlargement appears likely to be during the postnatal period and may begin as late as the latter part of the first year of life.  Enlargement during this period is consistent with either a decrease in the normal loss of neuronal processes (e.g., dendritic pruning) or over-exuberant dendritic arborization.  Our ability to link knowledge of brain patterning from genetic studies in animals (53) to existing data on patterns of brain development in autism will be important as candidate genes continue to be identified for autism (2).  The data we present on the possible timing of brain enlargement in autism also raise the possibility that the onset of autistic symptoms may be associated with postnatal changes in brain volume and that there may exist a ‘presymptomatic period’ in autism where intervention may have more potent effects. While this hypothesis is admittedly quite speculative, it is nevertheless important to consider.  Clearly the longitudinal study of very early brain and behavior development in autism has the potential to provide important clues relevant to early detection and early mechanisms underlying changes in the brain in autism.  

Several limitations of this study should be noted.  The difficulty ascertaining and scanning large numbers of non-autistic subjects at age two years, and particularly those with developmental delay without autism or other neurological sequelae often associated with developmental delay, limited our ability to have a larger comparison group for the autistic subjects in this study.  Also, typically developing children, scanned without the benefits of sedation, present challenges to obtaining high-quality MRI scans.  Given the young age of our autism group, we suspect that our sample may be biased towards more severely affected individuals who present clinically between 18-35 months of age. Certainly our sample of autistic subjects appears to have more mentally retarded individuals than appear in current population estimates of autism, where there is an  increasing prevalence of those in the higher IQ range (17).  We are currently enriching our sample of 4 year olds for higher IQ autistic individuals we were unable to find at age 2 years.  Despite these challenges, our study design attempts to take into account potential IQ effects by including both developmentally delayed and typically developing comparison children.

Acknowledgements

We express our appreciation for the assistance we received from the following:  the TEACCH centres, the NDRC Autism Subject Registry, and NC Children's Developmental Services Agencies for assisting with recruitment; Matthieu Jomier and Martin Styner for providing image processing support; Peg Nopoulos for her comments on the manuscript; and finally, and most importantly, the families who have participated in this study.  

References 

1.  American Psychiatric Association.  Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition. Washington, DC: American Psychiatric Association; 1994.

2.  Bailey A, LeCouteur A, Gottesman I, Bolton P, Simonoff E, Yudza E, Rutter M.  Autism as a strongly genetic disorder:  Evidence from a British twin study.  Psychol Med 1995; 25: 63-77. 

3.  Lainhart J, Piven J, Wzorek M, Santangelo S, Coon H, Folstein S:  Macrocephaly in children and adults with autism.  J Am Academy Child Adol Psychiatry. 1997; 36: 282-289.

4.  Stevenson R, Schroer R, Skinner C, Fender D, Simensen R.  Autism and macrocephaly.  Lancet. 1997; 349: 1744-1745.

5.  Bailey A, Luthert P, Bolton P, LeCouteur A, Rutter M. Autism and megalencephaly.  Lancet. 1993; 34: 1225-1226.

6.  Bauman M, Kemper T.  Histoanatomic observations of the brain in early infantile autism.  Neurology. 1985; 35: 866-874.

7.  Piven J, Nehme E, Simon J, Barta P, Godfrey P, Folstein S.  Magnetic resonance imaging in autism:  Measurement of the cerebellum, pons, and fourth ventricle.  Biol Psychiatry. 1992; 31: 491-504.

8.  Piven J, Ardnt S, Bailey J, Andreasen N.  Regional brain enlargement in autism:  A magnetic resonance imaging study.  J Am Acad Child Adol Psychiatry. 1996; 35: 530-536.

9.  Courchesne E, Karns C, Davis H, Ziccardi R, Carper R, Tigue Z, Chisum H, Moses P, Pierce K, Lord C, Lincoln A, Pizzo S, Schreibman L, Haas R, Akshoomoff N, Courchesne R.  Unusual brain growth patterns in early life in patients with autistic disorder:  An MRI study.  Neurology.  2001; 57: 245-254.

10.  Sparks B, Friedman S, Shaw D, Aylward E, Echelard D, Artru A, Maravilla K, Giedd J, Munson J, Dawson G, Dager S:   Brain structural abnormalities in young children with autism spectrum disorder.  Neurology. 2002; 59 (suppl 2): 184-92.

11.   Aylward EH, Minshew NJ, Field K, Sparks BF, Singh N.  Effects of age on brain volume and head circumference in autism.  Neurology.  2002; 59 (suppl 2): 175-183  .

12.  Herbert MR, Ziegler DA, Deutsch CK, O’Brien LM, Lange N, Bakardjiev A, Hodgson J, Adrien KT, Steele S, Makris N, Kennedy D, Harris GJ, Caviness, VS.  Dissociations of cerebral cortex, subcortical and cerebral white matter volumes in autistic boys.  Brain. 2003; 126, 1182-1192.

13.  Lotspeich LJ, Kwon H, Schumann CM, Fryer SL, Goodlin-Jones BL, Buonocore MH, Lammers CR, Amaral DG,  Reiss AL.  Investigation of neuroanatomical differences between autism and asperger syndrome.  Arch Gen Psychiatry. 2004; 61: 291-298.

14.  Giedd J, Blumenthal J, Jeffries N, Castellanos F, Liu H, Zijdenbos A, Paus T, Evans A, Rapoport J:  Brain development during childhood and adolescence:  a longitudinal MRI study.  Nat Neurosci 2001; 2: 861-863

15.  Torrey EF, Dhavale D, Lawlor JP, & Holden RH.  Autism and head circumference in the first year of life.  Biol Psychiatry.  2004; 56: 892-894.

16.  Courchesne E, Carper R, Akshoomoff N.  Evidence of brain overgrowth in the first year of life in autism.  JAMA.  2003; 290: 337-344.

17.  Fombonne E.  Epidemiological surveys of autism and other pervasive developmental disorders: an update.  J Aut Dev Disord. 2003; 33 (4): 365-382.

18.  Collaborative Linkage Study of Autism Group: Bradford Y, Haines J, Hutcheson H, Gardiner M, Braun T, Sheffield V, Cassavant T, Huang W, Wang K, Vieland V, Folstein S, Santangelo S, Piven J.   Incorporating Language Phenotypes Strengthens Evidence of Linkage to Autism. Am J Med Genet. 2001; 105:539-547.

19.  Piven J, Palmer P, Jacobi D, Childress D, Arndt S.  Broader autism phenotype:  evidence from a family history study of multiple-incidence autism families.  Am J Psychiatry. 1997; 154 (2): 185-190.

20.  Kuczmarski RG, Ogden CL, Guo SS, et al.  CDC growth charts for the United States:  Methods and development.  National Center for Health Statistics.  Vital Health Statistics, 2000; 11 (246), 1-187.

21.  Lord C, Rutter M, LeCouteur A. Autism Diagnostic Interview-Revised: a revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders. J Autism Dev Disord. 1994; 24(5): p. 659-85.

22.  DiLavore PC, Lord C, Rutter M.  The pre-linguistic autism diagnostic observation schedule. J Autism Dev Disord. 1995;  25(4): p. 355-79.

23.  Mullen EM.  Mullen Scales of Early Learning AGS Edition.  American Guidance Service, Inc.  USA; 1995.

24.  Sparrow SS, Balla DA, Cicche HV.  Vineland Adaptive Behaviour Scales-Interview Edition Survey Form Manual, Circle Pines: American Guidance Service, Inc; 1984.

25.  Zimmerman, IL, Steiner VG, Pond RE.  Preschool Language Scale-Fourth Edition (PLS-IV).  Psychological Corporation; 2002.

26.  Mesibov GB, Schopler E, Schaffer B, Michal N.  Use of the childhood autism rating scale with autistic adolescents and adults.  J Am Acad Child Adol Psychiatry. 1989; 28 (4): 538-541.

27.  Andreasen NC, Cohen G, Harris G, Cizadlo T, Parkinen J, Rezai K, Swayze VW II.  Image processing for study of the brain structure and function: Problems and programs. J Neuropsychiatry Clin Neurosci. 1992; 4: 125-133.

28.  Andreasen NC, Cizadlo T, Harris G, Swayze VW II, O'Leary DS, Cohen G, Ehrhardt J, Yuh WTC.  Voxel processing techniques for the antemortem study of neuroanatomy and neuropathology using magnetic resonance imaging. J Neuropsychiatry Clin Neurosci. 1993; 5: 121-130.

29.  Andreasen NC, Rajarethinan R, Cizadlo T, Arndt S, Swayze VW II, Flashman L, O'Leary DS, Ehrhardt J, Yuh WTC. Automatic atlas-based volume estimation of human brain regions from MR images. J Comput Assist Tomogr. 1996; 20: 98-106.

30.  Arndt S, Swayze V, Cizadlo T, O'Leary D, Cohen G, Yuh WTC, Ehrhardt JC, Andreasen NC.  Evaluating and validating two methods for estimating brain structure volumes:  Tessellation and simple pixel counting.  Neuroimage. 1994; 1: 191-198.

31.  Cohen G, Andreasen NC, Alliger R, Arndt S, Kuan J, Yuh WTC, Ehrhardt J.  Segmentation techniques for the classification of brain tissue using magnetic resonance imaging.  Psychiatry Research in Neuroimaging. 1992; 46: 33-51.

32.  Magnotta VA, Harris G, Andreasen NC, O'Leary DS, Yuh WT, Heckel D.  Structural MR image processing using the BRAINS2 toolbox. Comput Med Imaging Graph. 2002; 26: 251-264.

33.  Harris G, Andreasen NC, Cizadlo T, Bailey JM, Bockholt HJ, Magnotta VA, Arndt S. Improving tissue classification in MRI:  A three-dimensional multispectral discriminant analysis method with automated training class selection.  J Comput Assist Tomogr. 1999; 23 (suppl 1): 144-154.

34.  Van Leemput K, Maes F, Vandermeulen D, Suetens P.  Automated model-based bias field correction of MR images of the brain.  IEEE Transactions on Medical Imaging. 1999; 18: 885-896.

35.  Van Leemput K, Maes F, Vandermeulen D, Suetens P.  Automatedmodel-based tissue classification of MR images of the brain.  IEEE Transactions on Medical Imaging. 1999; 18: 897-908.

36.  Pastawa M, Gilmore J, Lin W, Gerig G. Automatic segmentation of neonatal brain MRI.  Lecture Notes in Computer Science (LNCS).  2004; 3216:  10-16.

37.  Park J, Gerig G, Chakos M, Vandermeulen D, Lieberman JA.  Neuroimaging of Psychiatry Disease: Reliable and efficient automatic brain tissue segmentation for increased sensitivity.  Schizophrenia Research.  2001; 49 (1-2):163.

38.  Styner M, Charles C, Park J, Gerig G.   Multisite validation of image analysis methods - Assessing intra and inter site variability.  Proc. SPIE.  2002;  4684: 278-286

39.  Hazlett HC, Gerig G, Smith RG, Jomier M, Jewells V, Piven J.  Development of an anatomical MRI atlas for two year old brain using high dimensional warping.  Submitted.

40.  Parent A.  Carpenter’s Human Neuroanatomy, 9th Ed.  Media, PA:  Williams & Wilkins; 1996.

41.  Crossman AR,  Neary D. Neuroanatomy: An Illustrated Colour Text. New York: Churchill Livingstone; 1995.

42.  Duvernoy H. The Human Brain: Surface Three-Dimensional Anatomy and MRI. New York: Springer-Verlag; 1991.

43.  Hanaway J, et al. The Brain Atlas: A Visual Guide to the Human Central Nervous System. Bethesda: Fitzgerald Science Press; 1998.

44.  Tamraz JC, Comair YG. Atlas of Regional Anatomy of the Human Brain Using MRI: With Functional Correlations. New York: Springer-Verlag; 2000.

45.  Mai JK, Assheuer J, Paxinos G. Atlas of the Human Brain. New York: Academic Press; 1997.

46.  Schnabel JA, Rueckert D, Quist M, Blackall JM, Castellano Smith AD, Hartkens T, Penney GP, Hall WA, Liu H, Truwit CL, Gerritsen FA, Hill DLG, Hawkes JD. A generic framework for non-rigid registration based on non-uniform multi-level free-form deformations. Lecture Notes in Computer Science. 2001; 2208: 573-581.

47.  Joshi S, Grenander U, Miller MI. On the geometry and shape of brain sub-manifolds. International Journal of Pattern Recognition and Artificial Intelligence: Special Issue on Processing of MR Images of the Human. 1997; 11 (8), 1317– 1343.

48.  Miller, MI, Joshi, SC, Christensen GE. Large deformation fluid diffeomorphisms for landmark and image matching.  In: Toga, Arthur W. (Ed.), Brain Warping, chapter 7. Academic Press; 1999.

49.  Joshi S, Davis B, Jomier M, Gerig G.  Unbiased Diffeomorphic Atlas Construction for Computational Anatomy.   NeuroImage; Supplement issue on Mathematics in Brain Imaging, (PM Thompson, MI Miller, T Ratnanather, RA Poldrack, and TE Nichols, eds.), vol. 23, no. Supplement 1, pp. S151-S160, Elsevier, Inc, 2004.

50.  Hazlett HC, Poe MD, Gerig G, Gimpel R, Piven J.  Cortical gray and white brain tissue volume in adolescents and adults with autism.  Submitted, under review, Biol Psych.

51.  Carper RA, Moses P, Tigue ZD, Courchesne E. Cerebral lobes in autism:  Early hyperplasia and abnormal age effects.  Neuroimage. 2002; 16:  1038-1051.

52.  Courchesne E, Carper RA, Akshoomoff, N:  Evidence of brain overgrowth in the first year of life in autism.  JAMA.  2004; 290: 337-344.

53.  Monuki ES, Walsh CA. Mechanisms of cerebral cortical patterning in mice andhumans. Nat Neurosci. 2001 Nov; 4 Suppl:1199-206.