Regional Brain Enlargement in Autism: A Magnetic Resonance Imaging Study

Regional Brain Enlargement in Autism: A Magnetic Resonance Imaging Study JOSEPH PIVEN , M .D. , STEPHA N ARNDT, PH.D ., JAM ES BAILEY, B.S., AN D NANCY AND REASEN , M .D., PII.D .

ABSTRACT Objective: To determine whether increased brain volume in autism, suggested in previous studies, is the result of general or regional brain size differences and to study the effect of gender on brain size and pattern of enlargement. Method: Total brain volume and cerebral cortical lobe volumes were examined in 35 autistic and 36 comparison subjects using magnetic resonance imaging and an automated method of brain volume measurement. Results: After controlling for height and nonverbal IQ, the authors detected a significant diagnosis x gender effect (F

= 7.4 : p = .009) for total

brain volume. A repeated-measures analysis of variance indicated that the pattern of enlargement (brain region x diagnosis) in autistic subjects differed from that in controls (F = 4.88; P

= .0004).

Subsequent sex-specific analysis

revealed significantly increased total brain volume in autistic males but not females. Analysis of lobe sizes showed significant enlargement in autistic subjects in temporal, parietal, and occipital, but not frontal lobes. Conclusions: These results suggest that brain size is increased in autism and that differences are not generalized but appear to be the result of a pattern of enlargement with increases in the size of specific cortical lobes. J. Am. Acad. Child Ado/esc. Psychiatry, 1996 , 35 (4):530-536. Key Words: autism, magnetic resonance imaging. brain, development.

er al., 1993) and head circum ference (Bailey et al., 1995; Steg and Rapop ort, 197 5; W alker, 1977 ) in aut ism. In a preliminary analysis of 22 autistic and 20 co ntrol male subjects exami ned with MRI , we recently found evidence of enlarged to tal brain (i.e., brain tissue volume + CSF volume), tot al tissue (i.e., brain volume - CSF), and total lateral ventricular volume in autistic males (Piven et al., 1995 ). T o explore the question of whether or not increased brain volume is a result of general or regional brain size differences and to study th e effect of gender on brain size and pattern of enlargeme nt, we report here th e analysis of MRI data from a larger sam ple of aut istic (26 male, 9 fem ale) and co ntrol (20 male, 16 fem ale) subjects usin g an automa ted method of brain volume measurement.

Struc tura l ima gin g studies have detected a variety of brain abno rma lities in autis m , but none of th ese find ings have been consiste ntly replicated (Mi nshew and D ombrowski, 1994). Several recent reports suggest the possibil ity th at hrain size in autis m may be increased. In a pre viou s magn etic reso nance im aging (M RI) study of posterior fossa struct ures, we reported significantly increased m idsagittal brain area in 15 high IQ aut istic males (Piven et al., 1992). Increased brain volume in autism was sim ilarly suggested in an abstract publ ication of MRI data from a study of 22 autistic subjects (Filipek et al., 1992 ). T hese M RI findings are co nsistent wi th reports of incre ased brain weight (Bailey

/ kuptet! AI/gl/st 23. 1995. From The Departm ent of Psychiatry, The University of Iotoa College of Med icine, IOlIJa City. This research was supported in part by a M arch OfDimes Clinical Research Gram (J. P.) and N IMH grants M H5/21 7 (J. P.) , M H OI 02 8 (J. P.) . M H40856 (N.A .), and M H4 327 1 (N. A.). l'lJe au th ors acknowledgethe helpfit! COl/III /nIlS of Robert Robinson. M.D .. in the preparation of this m anuscript. Rl'/' I'int requ ests to Dr. PiVCII, 1875 John Pappajolm Pa oilion, Department 0/ Psy chia try . University of Iowa H ospit als an d Clinics, Ioion Ci ty. lA 52242- 105 7. 0890 ·8567/ 96/.3504 -05.30$0.3 .00/0©1 9% by the American Academy

METHOD Sample T hi rty-five (26 ma le, 9 fema le) subjects who had p reviou sly received a diagnosis of auti stic disorder at the C hild Psych iatry C lin ic o f th e U niversity of Iowa H ospitals and Cli nics participated in this study. T h is sam ple included data fro m 22 autistic males reponed previously (Piven er al., 1995) . Subjects were selected for the presenr study if they were aged 12 years or older, likely to

or Child and Adolescent Psychiatry. 530

J.

AM . AC AD . CHILD !\DO LESC. PS YC H IATRY. 35:4, APRIl. 19 96

BRAIN FNI.ARGFMFNT IN AUTISM

complete a 20-minute MRI scan without requiring sedation, and had no history of a significant medical or neurological disorder. The mean age of the autistic subjects was 18.0 years (SD 4.5 years) (range: 12 through 29 years). Parental informants for all autistic subjects were interviewed with the Autism Diagnostic Interview (AD]) (LeCouteur et al., 1989). All subjects met both DSM-lII-R (American Psychiatric Association, 1987) and AD l-algorithrn (World Health Organization, 1992) criteria for autistic disorder. At the time of their initial evaluation in the Child Psychiatry Clinic, all autistic subjects received an unsrandardized physical examination that was reported to be without evidence of significant abnormality. Subjects were also tested with either the WAIS-R (Wechsler, 1981), the WISeR (Wechsler, 1991), or the Leiter International Performance Scales (Arthur, 1952) for measurement of nonverbal IQ. The mean IQ of the autistic subjects was 91.0 (SD 19.8) (range: 52 to 136). No autistic subject had a history of being treated for a seizure disorder. Around the time of their MRI scan, autistic subjects were examined for neurocutaneous markings of tuberous sclerosis and neurofibromatosis and measured for head circumference and height. The comparison group consisted of 36 healthy volunteers (20 males and 16 females) recruited from the community through newspaper advertisements. Twenty male controls had participated in a previous study from our group (Piven et al., 1995). Control subjects were a subset of an MRI database of the Mental HealthClinical Research Center (MH-CRC) at the University of Iowa, enriched for individuals with nonverbal IQ of 70 to 90 and for younger individuals (12 through 17 years). Comparison subjects were selected from the MH-CRC database who most closely resembled the age and IQ composition of the autistic subjects in this srudy. No comparison subject had a history of treatment for a psychiatric disorder (including alcohol or drug abuse or attentiondeficit hyperactivity disorder), history of a learning disability, or a significant medical or neurological disorder, as determined by structured interview (i.e., a modified version of the Comprehensive Assessment of Symptoms and History for control subjects) (Andreasen ct al., 1992b). Nonverbal IQ for all comparison subjects was assessed using the Performance subscales of the WAIS-R or WISC-Ill. The mean age of subjects in the comparison group was 20.2 years (SD 3.8 years) (range: 13 through 28 years) and the mean nonverbal IQ was 102.1 (SD 12.8) (range: 72 to 135).

MRI Acquisition Following explanation of the MRI procedure, informed consent was obtained from all subjects older than 17 years. Informed assent (as well as parental consent) was obtained from subjects younger than 18 years. MRI data were obtained with a T,-weighted threedimensional SPGR sequence on a 1.5-T scanner in the coronal plane using the following parameters: 1.5-mm slices with no gap; flip angle, 40°; repetition time (TR), 24 msec; echo time (TE) , 5 msec; two excitations; field of view, 26 cm; matrix, 256 X 192. This sequence yields approximately 124 contiguous slices through the entire brain and requires an acquisition time of approximately 20 minutes.

Image Analysis MRI data were processed by the locally developed family of software known as BRAINS (Andreasen et al., 1992a, 1993, 1994ac; Arndt er al., 1994; Cizadlo er al., 1994; Cohen et al., 1992) on a Silicon Graphics Personal Iris 4-D graphic work station, by a technician blind to the identity of the subjects. The various

J.

AM. ACAD. CHILD ADOUSC. PSYCHIATRY, 35:4, APRIl. 1996

components of this software have been validated using a variety of methods, including phantoms and postmortem tissue (Andreasen et al., I 992a, 1993, 1994b,c; Arndt et al., 1994; Cizadlo et al., 1994; Cohen cr al., 1992). Initially the whole brain was "cut out" of the skull by manually tracing along the pia-arachnoid junction, in consecutive slices to include the cerebral hemispheres, cerebellum, and brainsrcm down to the level of the vertebral arteries (i.c., the inferior boundary of the brainstern). The MRI data were then converted to a threedimensional data set using BRAIN BLAST, a voxcl processing program that does surface and volume rendering. Prior to subsequent analysis, the data were realigned and resampled in this threedimensional orientation to ensure comparability of head position across all subjects. At this stage of the processing, total brain (TBY) , total tissue, and total CSF volumes were generated. The total CSF volume was determined using an automated method of counting pixels that were "washed off," and it included CSF from the brain surface and from internal structures such as the ventricles (Arndt et al., 1994). Total tissue volume was also determined automatically by counting the remaining pixels; TBY was the simple sum of the other two variables. Intra- and intcrrarcr reliability for the measurement of TBY was high (intraclass correlation = .99 and .95, respectively). To obtain specific regional measurements in an automated manner, the volume-rendered brains were first rotated so that the anterior and posterior cornrnissurcs were in the same axial (x) and sagittal (y) planes. The interhemispheric fissure was then used to align the brains on the other axis (z}, Once in this standardized position, the brain was then divided into boxes, using a modification of the method first described by Talairach and Tournoux (1988), as follows. The distance between the midsagittal line and the extreme lateral margins of the brain is divided into four equalsized slices in the sagittal plane. The tissue in front of the anterior commissure (AC) is divided into four equal slices and the tissue behind the posterior commissure (PC) is divided into four equal slices, in the coronal plane. The tissue between the AC and PC is similarly divided into three equal slices. The tissue between the most superior point on the brain and the plane containing the AC and PC is then divided into four equal slices in the axial plane. To include the entire cerebellum, two slices were added to the inferior border of the standard Talairach atlas, producing a total of 1,232 stereotactically defined boxes. These boxes were then assigned, by experts in neuroanatomy, to one of 12 specific brain regions: left and right frontal, temporal, parietal, and occipital lobes and subcortical and cerebellar regions (Andreasen ct aI., 1994c; Talairach and Tournoux, 1988) (Fig. I). Boxes at the border of two regions of interest were assigned to the region of interest that had the majority of the box occupied with its tissue. Estimates of the cortical lobe volumes (frontal, temporal, parietal, and occipital lobes) produced by this automated method and used in this srudy have been compared with those produced by manual tracing in each of these specific regions in samples ranging from 20 to 56 subjects (Andreasen et al., 1994a,c; Talairach and Tournoux, 1988). The measurements were based on an anatomically precise simultaneous visualization of brain surface anatomy and of slices in three orthogonal planes, permitting a highly accurate delineation of the subregions (Fig. 2). The measurements produced by manual tracing were the criterion standard for assigning the accuracy of the automated measurements. The sensitivity of the automated measurements ranged from 82% to 93%, while the specificity was 98% to 99%. Because of the accuracy and efficiency of this automated method, it was used to measure regional brain volumes

531

rIVEN ET AI..

in cases and controls in this study. This method was reviewed in a previous study by Andreasen et a1. (1994a), using an MRI protocol identical with that of the present study.

Statistical Analysis In the present study, we use analysis of covariance (ANCOVA) to test for differences in TBV, after adjusting for height and

nonverbal IQ, and analysis of variance (AN OVA) to test for differences in specific lobe volumes between autistic and control subjects (SAS Institute Inc, 1992). Adjustments were made for nonverbal IQ, using ANCOVA, but we did not believe it was necessary to adjust for the minor but significant differences in mean ages of the cases and controls (18 and 20 years, respectively). This difference did not appear to be of any practical significance, and in a previous study (Piven et al., 1995) age was not correlated to brain size within the age range of our sample. A repeatedmeasures ANOVA was used to test for differences in the pattern of enlargement of lobes between autistic subjects and controls. Probability figures arc noted only if they achieved significance at conventional levels (p < .05).

RESULTS

Table 1 summarizes the results of the ANCOVA test of TBV adjusting for height (a measure of body size), nonverbal IQ, and gender in autistic and control subjects. In the total sample a significant diagnosis X gender effect was detected. Subsequent analyses showed a significant diagnosis effect in males. No significant

Fig. 1 The assignment of boxes in the stereotactic Talairach atlas to corresponding brain regions: frontal (orange), parietal (yellow), temporal (blue), occipital (purple), cerebellar (red), and subcortical (green). This permitted an automated atlas-based measurement of the volume-specific brain regions.

532

diagnosis effect was detected in females. Mean unadjusted TBV in autistic and control male subjects was 1,545.9 cc (SD = 149.7) and 1,437.9 cc (SD = 97.3); mean unadjusted TBV in autistic and control female subjects was 1,242.9 cc (SD = 175.7) and 1,303.1 cc (SD = 111.2). Inclusion of height and nonverbal IQ did not contribute significantly to any of the analyses. To examine the relationship between frontal, parietal, temporal, and occipital lobe volumes and to determine whether the pattern of enlargement in autism suggested a generalized or regional effect, we performed a repeated-measures AN OVA comparing the relationship between lobe volumes in autistic and comparison subjects. Results from this analysis demonstrated a significant region X diagnosis interaction (F = 4.88; df = 3,65; p = .0004) suggesting that enlargement was not generalized. This analysis also revealed a nonsignificant region X gender X diagnosis effect (F = 0.7; df = 3,65). Table 2 presents the data on mean lobe volumes in cases and controls and summarizes the results of the ANOVA comparing lobe volumes in these two groups.

Fig. 2 The accuracy of the automated measurements was assessed by comparing them to manually traced measurements of the specific regions. BRAINS software was used for three-dimensional volume rendering, which permitted visualization of the brain surface anatomy and internal structures in three orthogonal planes. For example, the central sulcus is shown here in orange on the surface rendering and its telegraphed coordinates appear in corresponding locations on the planar images. The area of the frontal lobe can then be accurately traced on individual slices (blue outline), and total slices can be summed to produce an estimate of total volume.

]. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 35:4, APRIL 1996

BRAIN ENLARGEMENT IN AUTISM

TABLE 1 Analysis of Covariance of Total Brain Volume, Controlling for Height and Nonverbal IQ in Autistic and Control Subjects Total

Diagnosis Height IQ Gender Dx X gender

Males

p"

p

pi>

0.11 0.01 1.36 35.03 7.44

NS NS NS .0001 .009

5.10 0.71 0.46

Females

P .03 NS NS

pc 2.86 0.90 0.78

P NS NS NS

Note: Dx = diagnosis; IQ = nonverbal IQ; NS = not significant.

"df = 4,57. /, dl= 1,39. cdf= 1,19.

The average volume of the temporal, parietal, and occipital lobes was significantly greater in autistic subjects than in controls. No significant group differences were noted in the average volume of the frontal lobes between the groups. The greatest average percent increase in size occurs in the occipital lobe (21.3%, or 24 cc), followed by the parietal (13.4%, or 29 cc) and temporal (7.7%, or 17 cc) lobes. Analysis of diagnosis X gender effect for the four lobes appears in Table 3, along with mean lobe volumes for male and female cases and controls. A significant diagnosis X gender interaction was noted for temporal, parietal, and frontal lobes. Post hoc analysis for sex-specific differences revealed that average volumes were significantly greater in male autistic versus male control subjects for both temporal (F = 12.58; df= 1,44; P < .0009) and parietal (F = 18.56; df = 1,44; P < .0001) lobes. The results regarding diagnosis (Table 2) and diagnosis X gender (Table 3) effects were consistent with nonparametric analyses, with the exception that no evidence of a significant diagnosis X gender interaction was detected for frontal lobe volume on nonparametric analysis.

a

DISCUSSION

In the present study, we demonstrate enlargement in TBV as well as increased volume of the parietal, temporal, and occipital lobes in autistic individuals. The results indicate that the TBV differences we detect are not simply a generalized phenomenon but that the relationship between volumes of the frontal, temporal, parietal, and occipital lobes differs between autistic individuals and the comparison group. These results are consistent with previous reports showing increased midsagittal brain area and brain volume on MRI (Filipek et aI., 1992; Piven et aI., 1992), increased head circumference (Bailey et aI., 1995; 5teg and Rapoport, 1975; Walker, 1977), and increased brain weight in autistic individuals (Bailey et al., 1995). The advantages of the automated method of brain size measurement used in this study have been described in detail in a recent report by Andreasen et al. (1993). This method provides an efficient and accurate way to obtain quantitative measurements of regional brain volumes. Volume measurements by the method we

TABLE 2 Comparison of Cortical Lobe Volumes in Autistic and Control Subjects Volume (cc) Autistic Subjects (n = 35)

Comparison Subjects (n = 36)

Lobe

Mean

SD

Mean

SD

p"

P

Temporal Parietal Occipital Frontal

238.1 248.9 139.3 408.9

31.0 38.5 33.0 68.5

221.2 219.5 114.8 404.0

21.3 28.00 20.1 14.9

4.68 10.85 11.50 0.54

.034 .002 .001 NS

Note: NS = not significant.

"
J.

AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 35:4, APRIl. 1996

533

PIYEN ET AI..

TABLE 3 Diagnosis X Cender Effect on Cortical Lobe Volume in Autistic and Control Subjects Volume (cc) Autistic Subjects Males (n ~ 26)

Control Subjects

Females (n ~ 9)

Males (n ~ 20)

Diagnosis X

Females (n ~ 16)

Gender

Lobe

Mean

SD

Mean

SD

Mean

SD

Mean

SD

F"

P

Temporal Parietal Occipital Frontal

250.6 262.3 149.3 435.1

20.7 28.8 29.7 40.5

201.8 210.3 110.4 333.3

27.7 37.9 24.3 78.4

229.1 228.7 121.3 424.7

20.2 22.4 17.6 40.8

211.3 208.1 106.6 378.2

18.8 30.6 20.5 40.3

8.30 4.52 3.80 5.36

.005 .037

NOlI':

"dl'~

NS ~ not significant.

1,67.

used are not subject to those biases inherent in using brain area measurements to estimate brain volumes. The finding in the study by Andreasen et al. (1994'1) of a decreased volume of the frontal lobes in schizophrenic individuals also supports the validity of this computerized method to detect meaningful dint~rences in the size of part icular brain regions. ()lII' finding of a pattern of brain eniargemelll' in autism suggests a more specific hypothesis lor [urure study than that sllggested by a more generali/.ed increase in brain volume. We do not yet know, however, the degree to which the abnormality we report is specific to aut ism. Filipek et .rl. (J (){)2) reported that brain volume in autism was greater than in subjects with developmelllal language disorder. Similarly, M RI studies of dyslexia (Schull'!. er al., J ()94) and attentionddicit hyperactivity disorder (Castellanos et ul., I ()94) suggest that incteased brain volume may not be common in other developmenr.il disorders. '1 'he results of this study suggest possible sex-specific differences in brain size in autism. The possibility that there are sex-specific differences in the brain in autism (i.e., males have enlarged brains a ill I females do not) would nOI be surprising. Sex-specific dint~rences in brain volume a III I brain physiology have been demon strated in uonuurisuc individuals (Schultz et al., 1994; Shaywil'!. er al., I ()9'i), and male brains are thought to be more vulnerable ro injury than brains of females (Auclreuseu el .il., I {)')o; Flaum et al., I ()()O). Sexspecific dif'lerences in the POPULll ion prevalence of aurism are also well known (LOIter, I ()6()), and sexspecific differences in the Inrnilia] aggregation of genet ically related disorders (Ritvo er al., J 989) and autistic phenomenology (Lord and ScllOpler, I ()H7) have also

S.H

NS

.024

been described. However, while the possibility of gender differences is consistent with findings in autism and other disorders, the alternative hypothesis of no gender difference must be considered. Recent data from a large epidemiological study of autism revealed evidence of increased head circumference in both males (n·c 70) and females (n = 21) with autism, with 2!t
I. AM. AUlD. UIIJ.I) AI)(}J.FSC. I'SYUIIATIZY, .1'):4, AI'IUJ. 19%

BRAIN ENLARGEMENT IN AUTISM

of an increase in neuronal density, suggesting the possibility that there may be an increase in the number of neurons in the brains of those individuals. Similar results were reported by Coleman et al. (1985). Sufficient data from neuropathological studies are not yet available to clarify which of these three possible mechanisms might underlie increased brain volume in autism. Further quantitative neuropathological examinations will be necessary to distinguish between the possible hypotheses we have suggested. Longitudinal studies of brain size and shape in autistic individuals may also offer insights into developmental mechanisms (e.g., increased neurogenesis versus decreased neuronal death) in the pathogenesis of autism. Further exploration of the findings suggested in this study may shed some light on the location of specific neuroanatomical abnormalities in autism. On the basis of our report of volume differences in all but the frontal lobes in autism, it is tempting to prematurely conclude that the abnormalities in autism are not in the frontal lobe. However, there are a number of reports suggesting the importance of frontal lobe deficits in understanding autism (Damasio and Mauer, 1978; Ozonoff et al., 1991; Zilbovicius et al., 1995), and several alternative hypotheses for explaining our findings should be considered. First, a simple, direct relationship between increased brain size and abnormal function cannot be assumed. Volume differences may only be indirectly related to the fundamental defects underlying autistic behavior. Second, our inability to detect significant size differences in the frontal lobe in autism cannot be taken to mean that no differences exist. The large size of the frontal lobes makes it more likely that regional size differences within the frontal lobe may have gone undetected. Third, the fact that the frontal lobes are the only lobes where enlargement was not detected in our autistic subjects might indicate that, relative to the size of the rest of the brain, it is the frontal lobes that should be viewed as most abnormal. Suppose, for example, that the genetic vulnerability to autism results in both enlarged brains as well as the familial aggregation of genetically related but milder aspects of the autism phenotype as described elsewhere (Piven and Folstein, 1994). However, for autistic disorder to develop in genetically vulnerable individuals, a second event must take place (e.g., the presence of a modifying gene or environmental event) that specifically affects (i.e., decreases) frontal lobe size and causes

]. AM. ACAD. CHILD ADOI.F.SC. PSYCHIATRY, .15:4, APRIl. 19%

the full syndrome of autism to result. Alternative hypotheses explaining the pattern of changes we detected should be considered before concluding the significance of these findings for understanding the neuroanatomical basis of autistic behavior. Finally, although these data do not point to a particular neuroanatomical deficit in autism, they suggest the possibility that brain enlargement may serve as a biological marker to define meaningful subgroups in this disorder. Possible correlates of increased brain volume including behavioral and cognitive parameters, craniofacial morphology, and possible etiological factors (e.g., perinatal factors and genes involved in brain development) should be explored in future studies. Previous reports of increased head circumference in autism together with the moderate correlation of head circumference with TBV (r = .67, P = .0001, in this sample) also suggest that head circumference may be a simple way to detect those with enlarged brain size in studies examining clinical correlates of increased brain size in autism. Nevertheless, continued study of the brain in autism, using more sophisticated image analysis techniques, is warranted and may contribute to our understanding of the complexities underlying the regional differences in brain size we have noted.

REFERENCES American Psychiatric Association (1 9il7), n;rlgnost;c rind Strll;st;ml MrIIl1It/1 oj"Mentfll Disorders, 3rd cd;t;on-rco;J£'t/ (nSM-III-lI). Washington, DC: American Psychiatric Association Andreasen NC, Cizadlo T, I Iarris C; ct al. (1993), Voxcl processing techniques f-()I" the antemortem study olncuroanaromy and ncuroparhology using magnetic resonance imaging. J N"Urop5yCIJ;fltlJ' en» Neuroscl 5: 12 1-130 Andreasen NC, Cohen G, Harris C; er al. (I 992a), Image processing for the study of brain structure and function: problems and programs . .I NfII/"Ops)'chirlt,] Clin Neurosci 4: 125-·133 Andreasen NC, Flashman L, l-laum M cr al. (1 994a), Regional brain abnormalities in schizophrenia measured with magnetic resonance imaging. .lAMA 272: 1763-1769 Andreasen NC, Flaum M, Arndt SV (1992b), The Comprehensive Assessmcnt of Symptoms and History (CASH): an instrument fill" assessing psychopathology and diagnosis. Arch (,"('IJ Psych;(/II] 49:515-623 Andreasen NC, Harris G, Cizudlo T cr al. (1994b), Techniques lill" measuring sulcal/gym! patterns in the brain as visualized through MR scanning: BRAINI'LOT and BRIAN MAl'. Proc Nmt Acad Sci U S

A 900:93-97 Andreasen NC, Rajaprabhakaran R, Cizadlo T, cds (I 994c), Automated Atlas-BasedDissection ofHuman BI"(/;"'/;"OIn MR flll(/ges. San Francisco, CA: Society of Magnetic Resonance in Medicine Andreasen NC, Swayze V, Flaum M, Yares WR, Arndt S, McChesney C (199()), Ventricular enlargement in schizophrenia evaluated with CT scanning: cHeers of gender, age, and stage of illness. Arch (,"l'II PsycbiaI'] 47: 1054-1 (J59

535

P IVEN ET AI..

Arndt S. Swayze V, C izadlo T et al. (I 994 ), Evaluating and validating two met ho ds fo r estimating bra in struc ture volumes: tessellation and sim ple pixel cou nt ing . Neuroimnge 1:191-19 8 Art hu r l; (195 2), The Arthur Adap tation ofthe Leiter International Perjormance Scal«. Ch icago: Psycho logical Ser vice C ent er Press Bailey A, LeCouteu r A. Gottesman I ct al, (I995) , Aut ism as a stro ng ly gen et ic disorder: eviden ce from a British twi n study. Psychol M ed 25 :63- 77 Bailey A, Lurh c rt 1'. Bolton 1'. LeC o uteu r A, Ru tter M (I 993 ). Aut ism an d mega lencep haly . Lancet .'14: 122 5- 1226 Bauma n M L, Kemper T L ( 1994), Neuroanat omi c o bservatio ns of th e brai n in autism . In: 17}{' Neurobiology of Aut ism, Bauma n M L, Kemper TL, cds , Baltimore: Joh ns H op kins Un iversiry Press, pp 119- 14 5 Ca stellanos XI' , Gie dd I N , Eckburg I' cr al. (I 994), Qua nt itati ve mo rp hology of the cau date nu cleus in att enti on defi cit hyperact ivity disorder. Am J Psychiatry 151 :179 1-1796 C izad lo T , Andreasen NC, Z eien G ct al, (I994) , Im age registration issues in the ana lysis of multiple-injection 150 H 20 PET st udies: BRAINFIT. Presente d at Society Photo-Optical Instr umentatio n Engineers, Internationa l Soc iety for Optical En gineering, 21 68:423- 430 Co hen G, And reasen NC, Alligcr R c t al. (1992) , Segment ation techniques for th e classification of brain tissue using m agne tic resonance imag ing. P,y chialry Res Neu roimaging 46 :33- 51 C oleman I'D , Romano J . Lapham I., Sim on W (1985) , Ce ll counts in ce reb ral co rtex of an auti stic patient . J Au tism Dell Disord 15:24 5- 255 Da masio AR , Ma uer RG (19 78), A neurologi cal model for childhood autism . Arch N euroI35:777- 86 3 Fili pek PA, Richelme C , Kennedy DN ct al . (19 92 ), Morphometric ana lysis of th e b rain in developmental lang uage disorders and autism (absrracr), A nn N mro132 :475 Flaum M , Arn d t S, And reasen N C (I990), T he role of gende r in studies of vent ricular enla rgeme nt in sch izop hr enia: a p redomin an tly ma le effect. Am J Psychiatry 147: 1327- 1332 LeC outeur A, Rutter M , Lord C et al. (1989) , Auti sm D iagno stic Int er view: a sta ndard ized investigator-b ased inst ru men t. J A utism D ell D isord 19:36 3-387 Lo rd C , Scho plcr E ( 1987), Neurobiologi cal imp lications of Sex differences in au tism . In : Neurobiological Issues in A utism, Scho plcr E, Me sibov BG , cds . N ew York : Plen um Press, pp 19 1- 21 1 Loner V ( 1966) , Epi de m iology of autistic co nd itio ns in yo ung child ren : I. Prevalence . Soc Psychiatry 1:124-1 37

536

Minshew NJ , Dombrowski SM (I994 ), In vivo neu roanato my of autism : ncuroimaging stud ies. In : Th e N eurobiology of Autism, Bauman ML, Kemper TL, cds , Baltimor e: Johns H o pkins University Press, p p 66- 85 O zo noff S, Pennington BI', Rogers SJ (199 1), Execut ive fu nction deficit s in h igh-fu nctioning auti st ic ind ivid ua ls: relationship to th eory of mind . J Child Psycbol Psychiatry 32 :108 1- 110 5 Pivcn J, Arndt S, Bailey J, H avercamp S, Andreasen N , Palmer P (19 95 ), An MRI st udy of brain size in aut ism . Am J Psychiatry 152 :1 14 5-1 149 Piven J, Folsrcin S (1994) , The gen etics o f autism . In: The Ne urobiology of Autism, Bauman ML, Kemper T L, cds . Balrimore: Joh ns H opkins Un iversity Press, pp 18-44 Piven J, Nehme E, Simo n J, Barta 1', Pearlson G, Folstcin SE (19 92) , Mag netic reson ance imagin g in autism : measuremen t o f th e cerebellum, po ns, and fou rt h vent ricle. BioI Psychiatry 3 1:49 1- 504 Rirvo Ek, Jorde LB, M ason -Bro thers A et al, ( 1989), The UC LA- U niversity of Utah ep idemiologic survey of autism : recu rre nce risk est imates and genetic co unseling. A m J Psychiatry 146 : 1032-1036 SAS Institu te Inc (1992), SAS/STAT User's Gui de, Version (6), 4th ed, Vo l 2 . Ca ry, NC: SAS Institute Inc Sch ultz RT, Cho NK, Staib LH et al, (199 4), Brain morphology in no rma l and dyslexi c chi ldre n : th e influence of sex and age. Ann N euroI35:732- 7 42 Shaywitz BA, Shaywirz SE , Pug h Kll. et al. (I995), Sex differences in the fun ctiona l organ izatio n of the bra in for language. Nature 373 :607-60 9 Steg J I', Rapoport JL (19 75) , Mi nor physical ano ma lies in norm al, neu ro tic learn ing di sabled , and severely disturbed ch ildr en . J Au tism Child Schizapbr 5:299-307 Talni rac h H , Tournoux 1', cds ( 198 8), Co-Planar Stereotaxic Atlas of the H U n/ail Brain. N ew York: Th iem e M edic al Publishers W alker H A (19 77 ), In ciden ce of minor ph ysical anomaly in autism . J Au tism Child Scbizopbr 7 :165 - 176 We chsler 0 (I 98 1), Wechsler Ad ult Intelligence Scale-Revised. San Antonio, T X: T he Psycho logical Corpo ration We chsler D (I9 91), W echsler Inte lligence Scale for Children -III . San Anto n io, T X: T he Psych ological C orporation W o rld Hea lth Organ ization (I992), C atego ries 1'00-1'99 mental an d bchavioural d isorde rs (incl uding disorders of psycho logica l development). C linica l descri ption s and diag no stic gu idelines. In : International Classification of Diseases, 10 reuision. G eneva: World H ealt h Orga nization Z ilbov icius M, Carreau B, Sam son Y er al. (I995) , Dela yed maturatio n of the frontal co rtex in child ho od auti sm . Am J Psychiatry 158:248-252

j , AM . AC AD . CHILD AD OL ESC . PSYCHI AT RY, 35:4, APRIL 19 96