Cerebral Cortex June 2010;20:1332--1340

doi:10.1093/cercor/bhp198

Advance Access publication October 9, 2009

Cortical Anatomy in Autism SpectrumDisorder: An In Vivo MRI Study on theEffect of Age

Armin Raznahan1, Roberto Toro2, Eileen Daly3,

Dene Robertson3, Clodagh Murphy3, Quinton Deeley3,

Patrick F. Bolton4, Tomas Paus2,5 and Declan G. M. Murphy3

1Department of Child and Adolescent Psychiatry, Institute of

Psychiatry, King’s College London, SE5 8AF, UK, 2Brain and

Body Center, University of Nottingham, NG7 2RD, UK,3Department of Psychological Medicine, Section of Brain

Maturation, Institute of Psychiatry, King’s College London,

SE5 8AF, UK, 4MRC Center for Social Genetic and

Developmental Psychiatry, Institute of Psychiatry, King’s

College London, SE5 8AF, UK and 5Montreal Neurological

Institute, McGill University, H3A 2B4, Canada

There is increasing evidence that children with autism spectrumdisorder (ASD) have age-related differences from controls in corticalvolume (CV). It is less clear, however, if these persist in adulthoodand whether these reflect alterations in cortical thickness (CT) orcortical surface area (SA). Hence, we used magnetic resonanceimaging to investigate the relationship between age and CV, CT, andSA in 127 males aged 10 through 60 years (76 with ASD and51 healthy controls). ‘‘Regional’’ analyses (using cortical parcellation)identified significant age-by-group interactions in both CV and CT (butnot SA) in the temporal lobes and within these the fusiform andmiddle temporal gyri. Spatially nonbiased ‘‘vertex-based’’ analysisreplicated these results and identified additional ‘‘age-by-group’’interactions for CT within superior temporal, inferior and medialfrontal, and inferior parietal cortices. Here, CV and CT were 1)significantly negatively correlated with age in controls, but not inASD, and 2) smaller in ASD than controls in childhood but vice versain adulthood. Our findings suggest that CV dysmaturation in ASDextends beyond childhood, affects brain regions crucial to socialcognition and language, and is driven by CT dysmaturation. This mayreflect primary abnormalities in cortical plasticity and/or besecondary to disturbed interactions between individuals with ASDand their environment.

Keywords: age, autism, brain, cortical thickness, MRI

Introduction

Autism spectrum disorder (ASD) is an increasingly recognized

(Baird et al. 2006) group of life span persistent neuro-

developmental disorders of early onset that are characterized

by abnormalities in language, social interaction, and a range of

stereotyped and repetitive behaviors (WHO 1993).

The brain basis of ASD is poorly understood. However,

longitudinal studies report that children with ASD have

accelerated growth in head circumference as compared with

controls (Hazlett et al. 2005; Dawson et al. 2007; Webb et al.

2007). In addition, many in vivo magnetic resonance imaging

(MRI) studies have reported that young children with ASD have

a significant increase in total brain volume (TBV) compared

with typically developing controls (e.g., see Hazlett et al. 2005).

In contrast, findings regarding differences in total brain volume

(TBV) between individuals with ASD and controls during later

childhood, adolescence, and adulthood have been less consis-

tent. Some groups report no group difference in TBV (e.g., see

Hallahan et al. 2008), and others find increased TBV in ASD

(e.g., see Hazlett et al. 2008). Given that findings of altered TBV

in ASD have been more consistently reported in pediatric

compared with adult populations, some have suggested that

people with ASD have differences in the trajectory of brain

growth that are limited to early childhood (for a meta-analysis,

see Redcay and Courchesne 2005).

Abnormal maturation of the cerebral cortex could contribute

to age-related differences in brain volume between people with

ASD and typically developing controls. For example, Courchesne

et al. (2001) compared children with autism to typically

developing controls aged between 2 and 16 years and reported

age-related differences in total cortical gray matter volume. In

a later study using the same sample, Carper et al. (2002)

compared lobar cerebral gray matter volume in children with

autism and controls. Their measure of cerebral gray matter

volume excluded subcortical structures and therefore reflected

cortical volume (CV). They found age-related differences in CV

between children with autism and controls that varied by lobe.

In the preschool years (2--4 years), children with autism had

significantly increased CV in the frontal and temporal lobes as

compared with controls. However, due to greater age-related CV

increase in controls compared with children with autism, these

differences were not apparent in older children. Also, there was

a significant ‘‘age-by-group’’ interaction in the CV of these

regions, and children with autism had a significantly smaller age-

related increase in CV than controls. Thus, prior work suggests

that in early childhood, people with ASD have significant

differences from controls in the maturation of CV. However, it

is unclear if these maturational differences normalize or persist

into adolescence and adulthood and if they vary by brain region.

We previously reported that adults (aged 18--45 years) with

ASD have significant age-related differences from controls in

whole-brain gray matter volume (those with Asperger

syndrome had a significantly smaller age-related reduction than

controls) (McAlonan et al. 2002). Our work provides pre-

liminary evidence that differences in brain maturation extend

beyond childhood. However, we did not directly measure CV

or include people from across the life span (i.e., children and

older adults).

Furthermore, CV has 2 sole determinants—cortical thick-

ness (CT) and surface area (SA). It is important to investigate

how abnormalities in CV relate to CT and SA because these

2 measures have distinct genetic determinants (Panizzoni et al.

2007) contrasting phylogeny (Rakic 1995) and differing

developmental trajectories (Armstrong et al. 1995; Sowell

et al. 2007). CT has been hypothesized to reflect dendritic

arborization/pruning within gray matter (Huttenlocher 1990)

or changing myelination at the gray/white matter interface

(Sowell et al. 2004), whereas SA is influenced by division of

� The Author 2009. Published by Oxford University Press. All rights reserved.

For permissions, please e-mail: journals.permissions@oxfordjournals.org

at NIH

Library on April 19, 2011

cercor.oxfordjournals.orgD

ownloaded from

progenitor cells in the embryological periventricular area

(Chenn and Walsh 2002) and varies as a function of brain

volume and cortical folding (gyrification). Establishing the

relative contribution of CT and SA disturbances could therefore

clarify the neurobiological mechanisms underlying CV abnor-

malities in people with ASD. New methods allow simultaneous

measurement of CV, CT, and SA from MRI scans (Fischl and

Dale 2000; MacDonald et al. 2000). These have not yet been

applied to examine age-related differences in CV between

people with ASD and controls or to establish the relative

contribution of CT and SA alterations to these. Hence, it is

unknown how abnormalities in CV relate to those in CT and SA,

which may reflect different neurobiological processes.

The methods currently available for examining the relative

contribution of CT and SA to alterations in CV require that

these 3 measures are taken from a set of predefined cortical

regions (that can be combined to give lobar estimates) (Fischl

et al. 2004). However, such ‘‘lobar’’ approaches to the study of

cortical maturation may be confounded because abnormalities

in neurodevelopment may alter traditional lobar boundaries

and/or extend across them. Therefore, a complementary

strategy is to assess cortical anatomy in a spatially nonbiased

manner by taking measures of CT at several thousand points

across the cortical sheet. To date, however, this approach has

never been applied to investigate cortical maturation in people

with ASD.

In summary, it has been proposed that young children with

ASD have age-related differences from controls in CV.

However, it is unclear if these normalize or persist into

adolescence and adulthood. Moreover, recent advances in MRI

data analysis make it possible to measure differences in the

2 determinants of CV (CT and SA)—which reflect very different

neurobiological processes. We therefore measured CV, CT, and

SA in people with ASD (n = 76) and controls (n = 51) aged

between 10 and 60 years. We tested the main hypothesis that

age-related differences in CV between people with ASD and

controls extend into adulthood. We also examined if matura-

tional differences in CV related to those in CT, SA, or both.

Finally, we carried out the first spatially nonbiased assessment

of age-related differences in CT between people with ASD and

controls in recognition of the fact that cortical abnormalities in

ASD may not conform to traditional lobar boundaries.

Materials and Methods

SampleWe included 127 male volunteers aged between 10 and 60 years: 76 with

ASD and 51 healthy controls. People with ASD were recruited through

a hospital-based national clinic and a university department specializing

in the assessment of ASD. The controls were drawn from the local

community through advertisement. Of those with ASD, 62 (82%) had

a diagnosis of Asperger syndrome, 10 (13%) autism, and 4 (5%) pervasive

developmental disorder not otherwise specified (PDD-NOS) or PDD

other. All were diagnosed clinically according to ICD-10 research criteria

(WHO 1993) by a team of senior clinicians trained in the autism

diagnostic interview revised (ADI-R) (Le Couteur et al. 1989) and the

autism diagnostic observation schedule (ADOS-G) (Lord et al. 2000). ASD

diagnoses were based on clinical interviews, collateral information from

family members, and review of other information available, such as

school reports. Forty-three (57%) of the participants, and/or their family

members, also agreed to confirmation of clinical diagnosis using the ADI-

R or ADOS. Thirty-three (43%) people were unable/unwilling to undergo

the ADI or ADOS. There were no significant demographic (age and

gender) or full-scale intelligence quotient (FSIQ) differences between

those with and without ADI/ADOS confirmation of ASD diagnosis.

Furthermore, there were no significant differences in mean age, FSIQ,

gender, or the proportion of participants with ADI/ADOS confirmation

of ASD diagnosis between ASD subgroups (i.e., Asperger vs. autism vs.

PDD-NOS). Despite this, we checked the robustness of our findings to

the exclusion of the 33 individuals with ASD who had been unable/

unwilling to undergo the ADI/ADOS.

All participants had FSIQs greater than 70. Participants in the study

underwent structured physical and psychiatric examination. No

participants had a history of neurological disorder affecting brain

function (e.g., epilepsy and stroke), major mental illness (e.g., psychosis,

major affective episode, or substance abuse), genetic disorder

associated with ASD (e.g., Fragile X syndrome), or a clinically abnormal

finding on MRI. No participants were taking psychotropic medication at

the time of the study. After complete description of the study to the

subjects, written informed consent was obtained.

Image Acquisition and AnalysisBrain MRIs were acquired using a GE Signa 1.5-T neuro-optimized MR

system (General Electric, Milwaukee, WI). ‘‘Freesurfer’’ freeware

(http://surfer.nmr.mgh.harvard.edu/fswiki) was used to derive models

of the cortical sheet in each T1-weighted image and parcellate out the

cortex into 35 regions. These well-validated (Fischl et al. 2004; Desikan

et al. 2006) and fully automated procedures have been extensively

described elsewhere (Fischl and Dale 2000; Fischl et al. 2004), and we

will only provide a brief description here. First, a single filled white

matter volume was generated for each cerebral hemisphere after

intensity normalization, the removal of extracerebral tissues using

a ‘‘skull stripping’’ algorithm, and image segmentation using a connected

components algorithm (Dale et al. 1999). Then, a surface tessellation

was generated for each white matter volume by fitting a deformable

template. The gray matter/cerebrospinal fluid (CSF) surface was also

modeled using a similar process. Given explicit models for the white/

gray and gray/CSF surfaces, the measure of absolute CT at any given

point on the white/gray matter surface was then taken to be the

shortest distance between that point and the gray/CSF surface.

This measurement was made at approximately 150 000 points across

each hemisphere for each scan.

Automated parcellation of each individual cortical hemispheric sheet

into 35 regions (Fischl et al. 2004) was achieved by aligning each scan

to a probabilistic atlas placed within a surface-based coordinate system.

Total CV, total SA, and average CT measures for these 35 subregions

were combined to generate lobar (frontal, temporal, parietal, and

occipital) estimates of total CV, total SA, and average CT.

For all analyses, CV was expressed as a proportion of TBV—corrected

cortical volume (cCV). Average TBV did not differ between cases and

controls nor was there significant age-by-group interaction for TBV. We

adjusted CV for TBV in order to remove variance in CV that was

associated with global differences in brain size and thus increase

statistical power to detect relationships between CV and our main

variables of interest (i.e., age, group, and their interaction).

Data Analysis

Analysis 1

Regional measures from automated parcellation of the cortical sheet.

In our first set of analyses, we used regional measures of cCV, CT, and

SA derived from automated parcellation of the cortical sheet. These

allowed us to examine the lobar and sublobar regional distribution of

age-related cCV difference in ASD and the interrelationship between

cCV, CT, and SA. We initially used multivariate analysis of variance

(MANOVA) to examine how variance in cCV, CT, and SA was explained

by a model including 1 between-group (group) and 2 within-group (age

and lobe) independent variables and their interaction terms. Next, for

those measures (i.e., cCV, CT, or SA) where significant variance was

explained by the ‘‘age-by-group-by-lobe’’ or age-by-group MANOVA

terms, we carried out post hoc analyses of age-by-group effects for that

measure within each lobe. Next, but only in those lobes where

a significant age-by-group effect for a cortical measure was found, we

examined age-by-group effects in each lobar subregion. Note that in all

Cerebral Cortex June 2010, V 20 N 6 1333

at NIH

Library on April 19, 2011

cercor.oxfordjournals.orgD

ownloaded from

these analyses, a significant main effect of ‘‘group’’ was not further

explored if it was accompanied by a significant age-by-group

interaction term. This is because the presence of a significant

interaction indicates that the effect of each term in the interaction is

dependent on the other and as such it is not coherent to consider the

main effect of each term in isolation. Also—the sole purpose of our

study was to investigate age-related group differences.

Analysis 2

A spatially nonbiased ‘‘vertex-based’’ approach. In our second set of

analyses, we examined the regional distribution of age-related differ-

ences in CT in people with ASD in a spatially nonbiased manner. This

additional analytic approach was adopted because cortical abnormal-

ities in ASD may well not conform to traditional lobar boundaries and in

order to provide the most spatially fine-grained analysis of CT

maturation in ASD available to date. At each of the approximately 150

000 points (vertices) in each hemisphere, the proportion of total

variance in CT accounted for by 2 linear regression models was

compared: One that did not include an age-by-group interaction term

[CT = intercept + (group 3 b1) + (age 3 b2)] and one that did [CT =intercept + (group 3 b1) + (age 3 b2) + (group 3 age 3 b3)]. This

resulted in an F ratio map for each hemisphere showing the degree to

which the inclusion of an age-by-group term increased the proportion

of CT variance that was accounted for.

The F ratio map was then thresholded using the false discovery rate

(FDR) method to correct for multiple comparison (Genovese et al.

2002)—with q (the proportion of rejected null hypotheses that would

be false rejections per family of tests) set at 0.05. We then examined

which of the regions also survived the more stringent FDR with q = 0.01.

A linear term for age was used for both ‘‘regional’’ and vertex-based

analyses. Higher order age terms (i.e., cubic and quadratic) were not

included because preliminary analyses indicated that in our cross-

sectional data set, the addition of these terms did not consistently

improve prediction of CV, CT, and SA. It is clear that large longitudinal

structural imaging data sets are required to properly model the linear,

quadratic, and cubic patterns of age-related change that occur in

different regions of the cortex (Shaw et al. 2008).

Results

Subject Characteristics

Subjects with ASD and controls did not differ in mean age.

There was, however, a significant group difference in FSIQ

(healthy controls had a significantly higher mean FSIQ than the

ASD group: P < 0.005). Please see Table 1 for full details.

Lobar and Sublobar Analyses of cCV, CT, and SA

MANOVA revealed a significant age-by-group-by-lobe interac-

tions for cCV (F = 3.4, P = 0.03, partial eta squared = 0.03) and

a significant age-by-group interaction for CT (F = 3.4, P = 0.03,

partial eta squared = 0.03) but not SA (P = 0.8) (see Table 2). We

therefore carried out post hoc lobar analysis of covariance

(ANCOVAs) for cCV and CT only (see Fig. 1). Significant age-by-

group effects were identified for temporal lobe cCV (F = 11.0,

P = 0.001, eta squared = 0.08) and CT (F = 6.6, P = 0.01, eta

squared = 0.05). We next examined within which temporal

lobe subregion this effect was most pronounced. ANCOVAs for

bilateral cCV and CT in each of the 9 temporal lobe subregions

revealed (after correction for multiple comparisons) significant

age-by-group effects for middle temporal cCV (F = 12.6, P =0.001, eta squared = 0.09) and CT (F = 16.3, P = 0.0005, eta

squared = 0.12) and for CT of fusiform gyrus (F = 8.3, P = 0.005,

eta squared = 0.06). Also, there was a trend toward significance

for fusiform cCV (F = 6.4, P = 0.01, eta squared = 0.05). Scatter

plots of cCV and CT against age for these temporal lobe

subregions revealed an identical pattern to that seen for total

temporal cCV and mean temporal CT. Controls showed age-

related reductions in cCV and CT, whereas in ASD, cCV and CT

were not significantly correlated with age. In younger

participants, cCV and CT were less in ASD than in controls,

but in older participants, this relationship was reversed. This

pattern of CV and CT decreases in ASD followed by increases

reflects the relative lack of age-related CT and cCV reduction in

people with ASD compared with controls.

The same pattern of age-by-group interactions for cCV and CT

being most marked within the temporal lobes held when we

limited our analysis to controls and the 43 people with ASD in

whom ADI-R or ADOS confirmation of diagnosis was available.

Plots of CT and CV for each lobe against age within this

subsample are available as supplementary material (Supplemen-

tary Figure 1). Furthermore, as in our larger sample, the temporal

lobe subregional measures showing most marked age-by-group

interactions were middle temporal cCV (F = 18.6, P = 0.0005)

and CT (F = 18.5, P = 0.0005) and fusiform cCV (F = 9.4, P =0.003) and CT (F = 6.9, P = 0.01). Our findings also held

when FSIQ was entered as a covariate. Age-related differences in

lobar cCV and CT between people with ASD and controls in

our data set do not reflect group differences in the variance of

these measures as assessed by Levene’s test for equality of

variances.

Table 1Subject characteristics

Characteristic Group

Cases (n 5 76) Controls (n 5 51)

Male 76 51Age

Mean 31.7 28.6Range 10--60 11--59Standard deviation 12.1 12.6

FSIQMeana 105 122Range 73--145 83--155Standard deviation 16.5 16.8

ASD diagnosisAsperger syndrome 62 —Autism 10 —PDD other 4 —

ADI/ADOS available 43 —

aSignificant difference P\ 0.005.

Table 2Multiple analysis of variance for cCV, CT, and SA

Term Measure

cCV CT SA

F P F P F P

Lobe 1733.3 0.0005 208.3 0.0005 1128.0 0.0005Age 97.0 0.0005 63.4 0.0005 4.8 0.03Group 4.7 0.03 5.2 0.03 0.3 0.5Age 3 lobe 49.3 0.0005 10.6 0.0005 3.9 0.03Group 3 lobe 3.0 0.04 2.4 0.09 0.4 0.6Group 3 age 4.4 0.04 2.8 0.1 0.3 0.6Group 3 age 3 lobe 3.4 0.03 3.4 0.03 0.2 0.8

Note: Significant test statistics are shown in bold. Note that SA across lobes does not vary as

a function of group. Group differences exist for lobar cCV and CT, but these vary as a function of

age and lobe.

1334 Cortical Maturation in ASD d Raznahan et al.

at NIH

Library on April 19, 2011

cercor.oxfordjournals.orgD

ownloaded from

Vertex-Based Analysis

Using this approach, we replicated the finding of the sublobar

analysis within the temporal lobe reported above. In addition,

highly significant age-related group differences in CT were

identified within the right (medial frontal, superior frontal,

inferior frontal, precentral, inferior temporal, parahippocampal

Figure 1. Lobar plots of cCV and CT against age for ASD and controls. Scatter plots for cCV are shown on the left and for CT on the right. Individuals with ASD (red) and controls(black) are shown separately. After application of Bonferroni correction for multiple comparisons, age-by-group interaction terms remain significant for temporal lobe cCV (F 511.0, P 5 0.001, eta squared 5 0.08) and CT (F 5 6.6, P 5 0.01, eta squared 5 0.05) only.

Cerebral Cortex June 2010, V 20 N 6 1335

at NIH

Library on April 19, 2011

cercor.oxfordjournals.orgD

ownloaded from

gyri, fusiform and inferior parietal lobule) and the left (medial

frontal, middle frontal, and superior temporal gyri) cortical

sheet. In all these regions, CT reduced as a function of age

among controls, but there was little or no relationship between

age and CT within the ASD group (see Table 3 and Figure 2).

Also, similar to the results from the ‘‘lobar analysis’’ above, in

younger participants, CT was less in ASD than in controls but in

older participants the relationship was reversed. These results

were much the same when we only included controls and

people with ASD in whom ADI-R or ADOS confirmation of

diagnosis was available (see Supplementary Figure 3 and

Supplementary Table 2). There remained regions of prominent

age-by-group interaction for CT within lateral temporal,

fusiform, and prefrontal cortices, and additionally peaks of

significant age-by-group interaction were identified within

right precuneus and superior temporal cortices. Also, age-by-

group interactions for mean CT within the identified cortical

regions remained highly significant with a model that included

FSIQ as a covariate.

Discussion

We compared maturation of the cerebral cortex and its

2 determinants (CT and SA) between people with ASD and

healthy controls aged from 10 to 60 years. We used 2 analytic

approaches—one which examined lobar and sublobar measures

of CV, CT, and SA derived from automated parcellation of the

cortical sheet and another which examined CT at several

thousand points across the cortical sheet in a spatially nonbiased

manner.

Both analytic techniques revealed that people with ASD have

significant age-related differences in cortical anatomy as

compared with controls. Our ‘‘regional analysis’’ identified

significant differences bilaterally in temporal lobe total cCV and

mean CT. Also within temporal lobe, those subregional

measures that differentiated ASD most from controls were

middle temporal cCV and CT and fusiform cCV. This finding

was replicated by the second spatially nonbiased analysis.

However, this latter approach also revealed significant age-by-

group interactions for CT in a number of other temporal,

frontal, and parietal cortical areas. All these regions demon-

strated the same phenomena: 1) in young people, cCV and CT

were increased in controls relative to ASD but 2) in controls

cCV and CT reduced significantly with increasing age, whereas

it did not in ASD, so that 3) by middle age/late adulthood cCV

and CT were greater in ASD compared with controls. In

contrast, SA showed minimal age-related differences within

either group, and this relationship did not differ across lobes or

between people with ASD and controls.

Hence, our results suggest that cortical dysmaturation in

ASD is not restricted to childhood but extends across the life

span. It is likely therefore that age is a very important variable

that may confound studies of brain anatomy (and particularly

those investigating subtle differences in cortical architecture)

in ASD. For example, our findings suggest that, a priori, studies

of CT that include only school-aged children will have very

different findings than those that include only adults (and vice

versa). To date, 3 published studies have compared CT in ASD

and controls—but without directly modeling the effects of age

(Chung et al. 2005; Hadjikhani et al. 2006; Hardan et al. 2006).

Two of these were carried out in youth, and the findings of one

are in keeping with what would be predicted from our study

(Chung et al. 2005), but the findings of the other (Hardan et al.

2006) are not. The third of these studies examined a relatively

small number of adults (14 cases and 14 controls) falling within

a wide age range (21--45 years) (Hadjikhani et al. 2006) and

identified regional CT reductions in ASD compared with

typically developing controls. Our data suggest that in the

absence of longitudinally characterized samples extending

across the life span, cross-sectional studies of CT in ASD will

not only need to be much larger than most of those carried out

to date but should also take into account the modulatory effect

of age.

In our sample, CT dysmaturation in ASD appeared to drive

age-related differences in CV between individuals with ASD and

typically developing individuals. Although the statistical signif-

icance of group (i.e., ASD vs. controls) differences in CV within

discreet age subgroups was not assessed, our data suggest that

in late childhood, CV within frontal and temporal lobes is

reduced in individuals with ASD compared with controls. This

is in apparent conflict with existing reports of brain volumetric

increases in ASD relative to controls in pediatric samples (Piven

et al. 1995; Courchesne et al. 2001; Sparks et al. 2002; Waiter

et al. 2004; Palmen et al. 2005; Hazlett et al. 2006; Lenroot

2008). However, many of these reports relate to whole-brain

volume (Piven et al. 1995; Sparks et al. 2002) or total gray

matter volume (CV and subcortical gray combined) (Waiter

et al. 2004; Palmen et al. 2005). It is possible for CV within

a given lobe to be reduced in individuals with ASD relative to

controls, whereas TBV or total gray matter volume shows no

significant group differences or significant increases in ASD.

There are 3 available studies of lobar CV in ASD (Carper et al.

2002; Hazlett et al. 2006; Lenroot 2008). These report CV

increases in preschool-aged children (Carper et al. 2002;

Lenroot 2008) and adolescents/young adults (Hazlett et al.

2006) with autism (including comorbid learning disabilities)

Table 3Cortical regions where age-by-group effects on vertex-based analyses exceeded correction for

multiple comparisons

Region Brodmannarea

Talairach coordinate

x y z

Left hemisphereMedial frontal gyrus 10 �9 49 �6Middle frontal gyrus 9 �34 21 26Superior temporal gyrus 41 �41 �39 6

22 �55 �51 13Middle temporal gyrus 21 �56 �23 213

21 �47 �1 �15Right hemisphere

Medial frontal gyrus 6 10 �11 62Superior frontal gyrus 8 6 29 50Precentral gyrus 6 45 �2 31Middle frontal gyrus 46 23 44 15

6 35 11 47Inferior frontal gyrus 46 39 34 5Middle temporal gyrus 21 53 �5 �22

21 56 �30 �1021 46 �49 422 58 �32 3

Inferior temporal gyrus 37 32 �68 �16Uncus 36 27 �4 �34Parahippocampal gyrus 34 19 0 �10Fusiform gyrus 20 43 �30 �23Inferior parietal lobule 40 45 �29 31

Note: All regions survived an FDR correction with q set at 0.05. Those regions in bold italics

survived correction with q 5 0.01.

1336 Cortical Maturation in ASD d Raznahan et al.

at NIH

Library on April 19, 2011

cercor.oxfordjournals.orgD

ownloaded from

relative to controls. Our data, however, suggest that CV is

decreased in adolescents with ASD (all without learning

disability and most with Asperger syndrome) compared with

controls. Therefore, our findings might differ from those of

others due to differences in the ASD subdiagnosis, cognitive

profile, and/or age range of the samples studied. With regard to

ASD subdiagnosis/cognitive profile, although it is clear that

differing clinical presentations within the autism spectrum can

be associated with common genetic risks (Bailey et al. 1995),

IQ range and ASD subdiagnosis have been shown to influence

reports of differences in cortical anatomy between people with

ASD and controls (McAlonan et al. 2008). With regards to age

range, CV increases in ASD relative to controls have been found

in both studies that cover the preschool years (Carper et al.

2002; Lenroot 2008). Our study and that of Hazlett et al. (2006)

both start in adolescence/early adulthood and generate

contrasting findings regarding CV alterations in ASD relative

to controls (globally increased CV and regionally decreased CV,

respectively). These findings would be consistent with a model

in which CV developmental trajectories in ASD and controls

are furthest apart in the preschool years but then converge

during childhood, making it harder to consistently identify

group differences. Large longitudinal studies will be required to

test this hypothesis.

Our study suggests a focal rather than global pattern of

cortical dysmaturation in ASD that is most pronounced in

regions that, within typically developing individuals, are

involved in social cognition (e.g., medial prefrontal, lateral

temporal, and lateral parietal cortices) (Lieberman 2007),

language (inferior frontal gyrus) (Price 2000), and some

aspects of executive function (dorsolateral prefrontal cortices)

(Salmon and Collette 2005)—domains in which many subjects

with ASD show impairments. Furthermore, in previous work,

these same cortical areas have been reported to be functionally

abnormal in children and adults with ASD during the

performance of a range of cognitive tasks involving ‘‘theory of

mind’’ processes (Castelli et al. 2002) (fusiform and superior

temporal regions), irony comprehension (Wang et al. 2006)

(medial prefrontal and lateral temporal regions), facial emotion

perception (Critchley et al. 2000; Dalton et al. 2005) (fusiform

region), set-shifting (Shafritz et al. 2008) (dorsolateral

prefrontal cortex), and tests of ‘‘central coherence’’ (Lee et al.

2007) (dorsolateral prefrontal cortex). In addition to functional

abnormalities, some of the cortical regions implicated in our

investigation have also been reported by postmortem studies to

show cellular abnormalities in ASD (e.g., in fusiform, middle

temporal, and dorsolateral prefrontal cortices) (Casanova et al.

2002; van Kooten et al. 2008). These overlaps raise the

Figure 2. Results of vertex-based analysis of age-by-group interactions across both cerebral hemispheres. Color maps show F ratio statistic with ‘‘warmer’’ colors indicatinga larger magnitude of age-by-group interaction. See Table 3 for details of those regions in which the F statistic exceeded the threshold imposed by FDR correction for multiplecomparisons. Scatter plots of mean CT against age for controls (solid line) and cases (dashed line) are shown for 2 of these regions.

Cerebral Cortex June 2010, V 20 N 6 1337

at NIH

Library on April 19, 2011

cercor.oxfordjournals.orgD

ownloaded from

possibility that atypical cortical maturation in ASD may partially

explain differences in cortical function (and/or vice versa).

Our combined analysis of CV, CT, and SA indicates that CV

dysmaturation in ASD is probably driven by differences in CT

rather than SA. Although relatively little is known of the factors

that shape age-related changes in CT during typical develop-

ment, interindividual differences in maturation of CT have been

related to IQ (Shaw et al. 2006), rate of improvement in

phonological processing (Lu et al. 2007), and single nucleotide

genetic polymorphisms (Shaw et al. 2007). Twin studies

suggest that at a group level, the relative role played by genetic

and environmental factors in CT variance differs both by

cortical region and age (Lenroot et al. 2009). The processes

influencing cortical maturation in ASD may be very distinct

from those that shape the typically developing cortex.

Therefore, we can only speculate as to what might underlie

atypical cortical maturation in ASD. One potential explanation

is that structural dysmaturation represents the direct influence

of genetic and environmental risk factors for ASD on cortical

plasticity. For example, genetically mediated abnormalities of

synaptic function in ASD (Garber 2007) could modulate

dendritic development and hence CT. Alternatively, dysmatu-

ration within a region could reflect the ‘‘secondary’’ impact that

growing up with ASD has on the developing brain. For example,

age-related reductions in CT may represent differences in

‘‘activity/experience-dependent’’ processes such as synaptic

pruning. Dysmaturation could therefore reflect ‘‘underuse’’ of

a cortical region that initially has normal developmental

potential. Candidate mechanisms for regional abnormalities of

cortical engagement in ASD include atypical patterns of eye

gaze to faces (Dalton et al. 2005). These 2 proposed accounts

for cortical dysmaturation in ASD raise distinct and testable

hypotheses.

Our study has several limitations. First, our study was cross-

sectional in design and thus susceptible to cohort effects in

which apparently age-related phenomena are an artifact of age

being nonrandomly distributed with respect to another variable

(e.g., IQ) that is associated with the outcome of interest (e.g.,

brain size). Of such potential confounders that were measured

in our study, age was not associated with FSIQ in our sample

nor were people with different ASD subdiagnoses differentially

distributed across the age range. Nevertheless, longitudinal

studies are central in the study of brain development—as

indicated by the dramatic interindividual variation and non-

linear growth trajectories identified by the few available large

longitudinal projects examining typical and atypical brain

maturation in humans (Giedd et al. 1999). Our pragmatic

design did, however, allow us to examine a very wide age range

(10--60 years) that would not be practically possible using

longitudinal approaches and represents the largest age range

studied to date in ASD using structural MRI. A second limitation

of our study is that confirmation of ASD diagnosis in our clinic-

based sample using ADI-R/ADOS was only possible in 57% of

cases. The validity of the ASD diagnoses made in the remaining

43% of cases is supported by the fact that multiple sources

of information were considered by a multidisciplinary team of

clinicians trained in ADI/ADOS before reaching a consensus

diagnosis. Moreover, our results did not change when we

excluded people without ADI/ADOS confirmation of clinical

ASD diagnosis. Finally, mean FSIQ in cases was significantly

lower than in controls. This reflects the fact that FSIQ

distribution within our controls was right-shifted relative to

the samples upon which IQ tests were standardized. This

phenomenon is also seen in some of the largest longitudinal

MRI data sets used to compare typically and atypically

developing populations (Shaw et al. 2006). Its potential impact

on our findings is lessened by the fact that current mean

population FSIQs are probably above 100 (Flynn 2000). Also,

the reduction in FSIQ typically seen in ASD complicates the

issue of group matching for FSIQ in ASD research (Szatmari

et al. 2004). Further, the possibility that our results were driven

by IQ differences between cases and controls rather than ASD

status per se is limited by the fact that 1) our findings remained

unaltered when we controlled for FSIQ and 2) the only

available longitudinal study relating FSIQ to cortical anatomy in

healthy controls (Shaw et al. 2006) found that maturational

differences in the cortices of groups defined by IQ were largely

restricted to frontal regions, whereas the age-related differ-

ences reported in our study are most marked in temporal

regions. Finally, as noted above, there was no significant group

difference in the relationship between age and IQ. Neverthe-

less, some cross-sectional studies (Narr et al. 2006) in healthy

controls have found an association between IQ and cortical

anatomy in regions that overlap with those identified in our

study, and as such, it is still conceivable that group differences

in IQ are a potential confound in our study. For the reasons

outlined above, however, we do not think that they can fully

explain the age-related differences in cortical anatomy that we

identified between individuals with ASD and controls.

In summary, we found that people with ASD have focal

rather than global age-related differences in cortical anatomy

from controls that extend into adulthood. Also, these are

mainly associated with variation in CT and not SA. This suggests

that, within people with ASD, differences in specific neurobi-

ological processes are not ‘‘fixed,’’ rather they may continue to

change across the life span.

Funding

UK Medical Research Council Clinical Research Training

Fellowship (A.R. to G0701370).

Supplementary Material

Supplementary material can be found at http://www.cercor.

oxfordjournals.org/

Notes

The authors wish to thank the MRC UK AIMS network for in-

frastructure support and the participants and their families for making

this study possible. Conflict of Interest : None declared.

Address correspondence to Dr Armin Raznahan. Email: armin.razna-

han@iop.kcl.ac.uk.

References

Armstrong E, Schleicher A, Omran H, Curtis M, Zilles K. 1995. The

ontogeny of human gyrification. Cereb Cortex. 5:56--63.

Bailey A, Le Couteur A, Gottesman I, Bolton P. 1995. Autism as

a strongly genetic disorder: evidence from a British twin study.

Psychol Med. 25:63--77.

Baird G, Simonoff E, Pickles A, Chandler S, Loucas T, Meldrum D,

Charman T. 2006. Prevalence of disorders of the autism spectrum in

a population cohort of children in South Thames: the Special Needs

and Autism Project (SNAP). Lancet. 368:210--215.

1338 Cortical Maturation in ASD d Raznahan et al.

at NIH

Library on April 19, 2011

cercor.oxfordjournals.orgD

ownloaded from

Carper RA, Moses P, Tigue ZD, Courchesne E. 2002. Cerebral lobes in

autism: early hyperplasia and abnormal age effects. Neuroimage.

16:1038--1051.

Casanova MF, Buxhoeveden DP, Switala AE, Roy E. 2002. Minicolumnar

pathology in autism. Neurology. 58:428--432.

Castelli F, Frith C, Happe F, Frith U. 2002. Autism, Asperger syndrome

and brain mechanisms for the attribution of mental states to

animated shapes. Brain. 125:1839--1849.

Chenn A, Walsh CA. 2002. Regulation of cerebral cortical size by

control of cell cycle exit in neural precursors. Science.

297:365--369.

Chung MK, Robbins SM, Dalton KM, Davidson RJ, Alexander AL,

Evans AC. 2005. Cortical thickness analysis in autism with heat

kernel smoothing. Neuroimage. 25:1256--1265.

Courchesne E, Karns CM, Davis HR, Ziccardi R, Carper RA, Tigue ZD,

Chisum HJ, Moses P, Pierce K, Lord C, et al. 2001. Unusual brain

growth patterns in early life in patients with autistic disorder: an

MRI study. Neurology. 57:245--254.

Critchley HD, Daly EM, Bullmore ET, Williams SCR, van Amelsvoort T,

Robertson DM, Rowe A, Phillips M, McAlonan G, Howlin P, et al.

2000. The functional neuroanatomy of social behaviour: changes in

cerebral blood flow when people with autistic disorder process

facial expressions. Brain. 123:2203--2212.

Dale AM, Fischl B, Sereno MI. 1999. Cortical surface-based analysis. I.

Segmentation and surface reconstruction. Neuroimage. 9:

179--194.

Dalton KM, Nacewicz BM, Johnstone T, Schaefer HS, Gernsbacher MA,

Goldsmith HH, Alexander AL, Davidson RJ. 2005. Gaze fixation and

the neural circuitry of face processing in autism. Nat Neurosci.

8:519--526.

Dawson G, Munson J, Webb SJ, Nalty T, Abbott R, Toth K. 2007. Rate of

head growth decelerates and symptoms worsen in the second year

of life in autism. Biol Psychiatry. 61:458--464.

Desikan RS, Segonne F, Fischl B, Quinn BT, Dickerson BC, Blacker D,

Buckner RL, Dale AM, Maguire RP, Hyman BT, et al. 2006. An

automated labeling system for subdividing the human cerebral

cortex on MRI scans into gyral based regions of interest. Neuro-

image. 31:968--980.

Fischl B, Dale AM. 2000. Measuring the thickness of the human cerebral

cortex from magnetic resonance images. Proc Natl Acad Sci USA.

97:11050--11055.

Fischl B, Van Der KA, Destrieux C, Halgren E, Segonne F, Salat DH,

Busa E, Seidman LJ, Goldstein J, Kennedy D, et al. 2004.

Automatically parcellating the human cerebral cortex. Cereb

Cortex. 14:11--22.

Flynn JR. 2000. IQ gains, WISC subtests and fluid g: g theory and the

relevance of Spearman’s hypothesis to race. Novartis Found Symp.

233:202--216.

Garber K. 2007. Neuroscience. Autism’s cause may reside in abnormal-

ities at the synapse. Science. 317:190--191.

Genovese CR, Lazar NA, Nichols T. 2002. Thresholding of statistical

maps in functional neuroimaging using the false discovery rate.

Neuroimage. 15:870--878.

Giedd JN, Blumenthal J, Jeffries NO, Castellanos FX, Liu H, Zijdenbos A,

Paus T, Evans AC, Rapoport JL. 1999. Brain development during

childhood and adolescence: a longitudinal MRI study. Nat Neurosci.

2:861--863.

Hadjikhani N, Joseph RM, Snyder J, Tager-Flusberg H. 2006. Anatomical

differences in the mirror neuron system and social cognition

network in autism. Cereb Cortex. 16:1276--1282.

Hallahan B, Daly EM, McAlonan G, Loth E, Toal F, O’Brien F,

Robertson D, Hales S, Murphy C, Murphy KC, et al. 2008. Brain

morphometry volume in autistic spectrum disorder: a magnetic

resonance imaging study of adults. Psychol Med. 39:337--346.

Hardan AY, Muddasani S, Vemulapalli M, Keshavan MS, Minshew NJ.

2006. An MRI study of increased cortical thickness in autism. Am J

Psychiatry. 163:1290--1292.

Hazlett HC, Poe M, Gerig G, Smith RG, Provenzale J, Ross A, Gilmore J,

Piven J. 2005. Magnetic resonance imaging and head circumference

study of brain size in autism: birth through age 2 years. Arch Gen

Psychiatry. 62:1366--1376.

Hazlett HC, Poe MD, Gerig G, Smith RG, Piven J. 2006. Cortical gray and

white brain tissue volume in adolescents and adults with autism.

Biol Psychiatry. 59:1--5.

Huttenlocher PR. 1990. Morphometric study of human cerebral cortex

development. Neuropsychologia. 28:517--527.

Le Couteur A, Rutter M, Lord C, Rios P. 1989. Autism diagnostic

interview: a standardized investigator-based instrument. J Autism

Dev Disord. 19:363--387.

Lee PS, Foss-Feig J, Henderson JG, Kenworthy LE, Gilotty L,

Gaillard WD, Vaidya CJ. 2007. Atypical neural substrates of

embedded figures task performance in children with autism

spectrum disorder. Neuroimage. 38:184--193.

Lenroot R. 2008. Increased cortical thickness and gray matter volume in

young children with autism. In: International Meeting for Autism

Research; 2008, May 15, London.

Lenroot RK, Schmitt JE, Ordaz SJ, Wallace GL, Neale MC, Lerch JP,

Kendler KS, Evans AC, Giedd JN. 2009. Differences in genetic and

environmental influences on the human cerebral cortex associated

with development during childhood and adolescence. Hum Brain

Mapp. 30:163--174.

Lieberman MD. 2007. Social cognitive neuroscience: a review of core

processes. Annu Rev Psychol. 58:259--289.

Lord C, Risi S, Lambrecht L, Cook EH, Jr, Leventhal BL, DiLavore PC,

Pickles A, Rutter M. 2000. The autism diagnostic observation

schedule-generic: a standard measure of social and communication

deficits associated with the spectrum of autism. J Autism Dev

Disord. 30:205--223.

Lu L, Leonard C, Thompson P, Kan E, Jolley J, Welcome S, Toga A,

Sowell E. 2007. Normal developmental changes in inferior frontal

gray matter are associated with improvement in phonological

processing: a longitudinal MRI analysis. Cereb Cortex. 17:1092--1099.

MacDonald D, Kabani N, Avis D, Evans AC. 2000. Automated 3-D

extraction of inner and outer surfaces of cerebral cortex from MRI.

Neuroimage. 12:340--356.

McAlonan GM, Daly E, Kumari V, Critchley HD, van AT, Suckling J,

Simmons A, Sigmundsson T, Greenwood K, Russell A, et al. 2002.

Brain anatomy and sensorimotor gating in Asperger’s syndrome.

Brain. 125:1594--1606.

McAlonan GM, Suckling J, Wong N, Cheung V, Lienenkaemper N,

Cheung C, Chua SE. 2008. Distinct patterns of grey matter

abnormality in high-functioning autism and Asperger’s syndrome. J

Child Psychol Psychiatry. 49:1287--1295.

Narr KL, Woods RP, Thompson PM, Szeszko PR, Robinson D,

Dimtcheva T, Gurbani M, Toga AW, Bilder RM. 2006. Relationships

between IQ and regional cortical gray matter thickness in healthy

adults. Cereb Cortex. 17:263--271.

Palmen SJ, Hulshoff Pol HE, Kemner C, Schnack HG, Durston S,

Lahuis BE, Kahn RS, van EH. 2005. Increased gray-matter volume in

medication-naive high-functioning children with autism spectrum

disorder. Psychol Med. 35:561--570.

Panizzoni CF-N, Eyer L, Jerningan T, Prom-Wormley E, Neale M,

Jacobsen K, Lyons MJ, Grant MD, Franz CE, Xian H, et al. 2007.

Distinct genetic influence on cortical surface area and cortical

thickness. Cereb Cortex. Advance Access published March 18,

doi:10.1093/cercor/bhp026.

Piven J, Arndt S, Bailey J, Havercamp S, Andreasen NC, Palmer P. 1995.

An MRI study of brain size in autism. Am J Psychiatry.

152:1145--1149.

Price CJ. 2000. The anatomy of language: contributions from functional

neuroimaging. J Anat. 197:335--359.

Rakic P. 1995. A small step for the cell, a giant leap for mankind:

a hypothesis of neocortical expansion during evolution. Trends

Neurosci. 18:383--388.

Redcay E, Courchesne E. 2005. When is the brain enlarged in autism?

A meta-analysis of all brain size reports. Biol Psychiatry. 58:

1--9.

Salmon E, Collette F. 2005. Functional imaging of executive functions.

Acta Neurol Belg. 105:187--196.

Shafritz KM, Dichter GS, Baranek GT, Belger A. 2008. The neural

circuitry mediating shifts in behavioral response and cognitive set in

autism. Biol Psychiatry. 63:974--980.

Cerebral Cortex June 2010, V 20 N 6 1339

at NIH

Library on April 19, 2011

cercor.oxfordjournals.orgD

ownloaded from

Shaw P, Gornick M, Lerch J, Addington A, Seal J, Greenstein D, Sharp W,

Evans A, Giedd JN, Castellanos FX, et al. 2007. Polymorphisms of the

dopamine d4 receptor, clinical outcome, and cortical structure in

attention-deficit/hyperactivity disorder. Arch Gen Psychiatry.

64:921--931.

Shaw P, Greenstein D, Lerch J, Clasen L, Lenroot R, Gogtay N, Evans A,

Rapoport J, Giedd J. 2006. Intellectual ability and cortical de-

velopment in children and adolescents. Nature. 440:676--679.

Shaw P, Kabani NJ, Lerch JP, Eckstrand K, Lenroot R, Gogtay N,

Greenstein D, Clasen L, Evans A, Rapoport JL, et al. 2008.

Neurodevelopmental trajectories of the human cerebral cortex. J

Neurosci. 28:3586--3594.

Sowell ER, Peterson BS, Kan E, Woods RP, Yoshii J, Bansal R, Xu D,

Zhu H, Thompson PM, Toga AW. 2007. Sex differences in cortical

thickness mapped in 176 healthy individuals between 7 and 87

years of age. Cereb Cortex. 17:1550--1560.

Sowell ER, Thompson PM, Leonard CM, Welcome SE, Kan E, Toga AW.

2004. Longitudinal mapping of cortical thickness and brain growth

in normal children. J Neurosci. 24:8223--8231.

Sparks BF, Friedman SD, Shaw DW, Aylward EH, Echelard D, Artru AA,

Maravilla KR, Giedd JN, Munson J, Dawson G, et al. 2002. Brain

structural abnormalities in young children with autism spectrum

disorder. Neurology. 59:184--192.

Szatmari P, Zwaigenbaum L, Bryson S. 2004. Conducting genetic

epidemiology studies of autism spectrum disorders: issues in

matching. J Autism Dev Disord. 34:49--57.

van Kooten I, Palmen SJ, von CP, Steinbusch HW, Korr H, Heinsen H,

Hof PR, van EH, Schmitz C. 2008. Neurons in the fusiform gyrus are

fewer and smaller in autism. Brain. 131:987--999.

Waiter GD, Williams JH, Murray AD, Gilchrist A, Perrett DI, Whiten A. 2004.

A voxel-based investigation of brain structure in male adolescents with

autistic spectrum disorder. Neuroimage. 22:619--625.

Wang AT, Lee SS, Sigman M, Dapretto M. 2006. Neural basis of irony

comprehension in children with autism: the role of prosody and

context. Brain. 129:932--943.

Webb SJ, Nalty T, Munson J, Brock C, Abbott R, Dawson G. 2007. Rate of

head circumference growth as a function of autism diagnosis and

history of autistic regression. J Child Neurol. 22:1182--1190.

WHO 1993. Mental disorders: a glossary and guide to their classification

in accordance with the 10th revision of the international

classification of diseases-research diagnostic criteria (ICD-10).

Geneva (Switzerland): WHO.

1340 Cortical Maturation in ASD d Raznahan et al.

at NIH

Library on April 19, 2011

cercor.oxfordjournals.orgD

ownloaded from