Compared to What? Early Brain Overgrowth in Autism and the Perils of Population Norms

ARCHIVAL REPORT

Compared to What? Early Brain Overgrowth inAutism and the Perils of Population Norms

Background: Early brain overgrowth (EBO) in autism spectrum disorder (ASD) is among the best replicated biological associations inpsychiatry. Most positive reports have compared head circumference (HC) in ASD (an excellent proxy for early brain size) with well-known reference norms. We sought to reappraise evidence for the EBO hypothesis given 1) the recent proliferation of longitudinal HCstudies in ASD, and 2) emerging reports that several of the reference norms used to define EBO in ASD may be biased toward detectingHC overgrowth in contemporary samples of healthy children.

Methods: Systematic review of all published HC studies in children with ASD. Comparison of 330 longitudinally gathered HC measuresbetween birth and 18 months from male children with autism (n ¼ 35) and typically developing control subjects (n ¼ 22).

Results: In systematic review, comparisons with locally recruited control subjects were significantly less likely to identify EBO in ASDthan norm-based studies (p � .001). Through systematic review and analysis of new data, we replicate seminal reports of EBO in ASDrelative to classical HC norms but show that this overgrowth relative to norms is mimicked by patterns of HC growth age in a largecontemporary community-based sample of US children (n � 75,000). Controlling for known HC norm biases leaves inconsistent supportfor a subtle, later emerging and subgroup specific pattern of EBO in clinically ascertained ASD versus community control subjects.

Conclusions: The best-replicated aspects of EBO reflect generalizable HC norm biases rather than disease-specific biomarkers. Thepotential HC norm biases we detail are not specific to ASD research but apply throughout clinical and academic medicine.

Key Words: Autism, bias, CDC, head circumference, systematicreview, WHO

An atypical pattern of brain growth during early postnatal lifewas first reported in children with autism over a decadeago (1). Since then, numerous studies have used head

circumference (HC) measures or in vivo structural magnetic reso-nance imaging estimates of brain size to test for early brainovergrowth (EBO) in children with autism spectrum disorder (ASD).

The EBO hypothesis, which states that ASD is associated withan abnormal acceleration of brain growth within the first 2 yearsof life (2), has received considerable empirical support, leading tothe speculation that EBO might be a potential biomarker for ASD(3). The influence of EBO reports on ASD research is evidenced byrecent use of the link between brain enlargement and ASD tovalidate or interpret 1) animal models for putative genetic (4) andepigenetic (5) risk mechanisms in ASD; 2) studies of postmortembrain tissue from individuals with ASD (6,7); 3) reported associ-ations between a given genetic variant and risk for ASD (8); and4) in vivo neuroimaging and electrophysiological studies of

From the Child Psychiatry Branch (AR, DG, JNG) and the Laboratory of Brainand Cognition (GLW, LA, AM), National Institute of Mental Health,National Institutes of Health, Bethesda, Maryland; Department ofPsychiatry (RL), University of New South Wales, Sydney, Australia;Pediatric Developmental Neurosciences Branch (AT, MG, SES), NationalInstitute of Mental Health, National Institutes of Health, Bethesda,Maryland; and Department of Neurology (SS), Children’s Hospital Boston,Harvard Medical School, Boston, Massachusetts.

Address correspondence to Armin Raznahan, M.D., Ph.D., National Instituteof Mental Health, National Institutes of Health, Child Psychiatry Branch,Room 4C108, Building 10, 10 Center Drive, Bethesda, MD 20892;E-mail: [email protected].

Received Nov 27, 2012; revised Feb 27, 2013; accepted Mar 13, 2013.

0006-3223/$36.00http://dx.doi.org/10.1016/j.biopsych.2013.03.022 Publis

altered brain connectivity in ASD (9–11). Two recent develop-ments urge reappraisal of the evidence base for EBO in ASD,however.

First, several new longitudinal studies of early brain growth inASD have become available since the topic last underwentsystematic review (12). Longitudinal data are critical for testingthe EBO hypothesis, which hinges on the presence of an atypicalpattern of brain size change in ASD (13). Currently, the largestavailable body of evidence regarding early brain growth in ASDcomes from studies of HC, which serves as an excellent proxy forbrain size in infants and preschool-aged children (14,15) andprovides cost-effective access to large bodies of retrospectivelongitudinal data about brain growth in ASD during the first yearsof postnatal life. There are now 11 longitudinal HC studiesof brain growth in ASD within the hypothesized phase of EBO(15–25), which together provide 17 times (�3000:180) moreobservations than the two existent longitudinal structural neuro-imaging studies of preschoolers with ASD (26,27). As 10 of these11 longitudinal HC studies have been published since the topic ofEBO was last systematically reviewed (12), there is a pressingneed to formally integrate the now much expanded evidencebase regarding patterns of early brain growth in ASD. Suchintegration could also help clarify recently posed questionsregarding the selectivity of EBO for certain ASD subgroups [e.g.,as defined by sex or clinical status (22)] and the extent to whichEBO in ASD is part of more generalized somatic overgrowth (21).

The second recent development that could significantlymodify our understanding of EBO in ASD comes from multiplestudies outside the field of ASD, which report discrepanciesbetween HC growth reference norms commonly used to testthe EBO hypothesis in ASD and contemporary patterns of HCgrowth (28–33). The best replicated of these discrepanciesconcerns Center for Disease Control and Prevention (CDC) norms(34): to date, five large independent contemporary samples ofhealthy children have been reported to show trajectories of HCgrowth during the first year of life that are abnormally accelerated

BIOL PSYCHIATRY 2013;74:563–575hed by Elsevier Inc on behalf of Society of Biological Psychiatry

564 BIOL PSYCHIATRY 2013;74:563–575 A. Raznahan et al.

relative to the CDC norms (29–33). This discrepancy is strikinglysimilar to the pattern of HC overgrowth relative to CDC norms thatconstitutes the principal finding in five of the most commonly citedsources of empirical support for the EBO hypothesis in ASD(15,16,19,20,35). This convergence raises a critical question—mightreports of HC growth in ASD that diverge from CDC norms reflect asystematic bias in CDC norms rather than any specific associationbetween ASD and an abnormal acceleration of early brain growth?Existing studies of bias in CDC norms point towards possiblelimitations in HC data collection and modelling (30). Concerns alsoarise regarding potential biases in other popular HC norms, givenrobust and convergent evidence that the tempo of human HCgrowth can show robust intergenerational change and that suchsecular changes in HC growth can emerge within time frames thathave commonly separated the construction of these HC norms andtheir application in ASD research [United Kingdom (28,36), China(37), Netherlands (38), Finland (32), Japan (39), and Korea (40)]. Theinfluence of HC norm use on early brain growth findings in ASD isyet to be systematically assessed however, despite carryingsignificant consequences for the EBO hypothesis in ASD, and moregenerally, for the application of population reference norms inclinical and academic medicine.

In the present report, we reappraise evidence for EBO in ASDthrough an updated systematic review of all HC studies inchildren with ASD and analysis of HC data from a recentlyassembled cohort of children with autism and typically devel-oping control subjects (41). Our study builds on the lastsystematic review of the EBO hypothesis (12) in two main ways.First, we incorporate the expanded evidence base regardingEBO in ASD, which has been enriched by multiple longitudinalstudies capable of quantifying those changes in brain size thatare so critical to the EBO hypothesis. Second, we provide adetailed methodological annotation of published research, sothat available evidence for and against the EBO hypothesis canbe considered in light of pertinent study features including1) age range; 2) temporal density of measures used to generategroup HC estimates; 3) extent of control for potential HCmodifiers such as body size, sex, and ethnicity; and 4) sourceof control data with which ASD HC measures are contrasted.

The relationship between control group selection and HCfindings in ASD is a major empirical focus in both our systematicreview and analysis of new HC data, which we address by1) comparing EBO findings in ASD studies that rely exclusively onHC norms with ASD studies that include locally recruited controlchildren as a comparison group; 2) comparing EBO findings in ASDacross multiple HC reference norms; and 3) using the commonreference frame of CDC norms to integrate reported patterns ofearly HC growth in children with ASD and locally recruited controlsubjects and then compare these trajectories with the largest(�400,000 HC measures on �75,000 children ages 0 to 18 months)and most current description of HC growth in a community-basedsample of US children (Primary Care Network [PCN] norms) (30).

Methods and Materials

Systematic ReviewThree authors (A.R., G.L.W., L.A.) independently carried out an

electronic literature search (PubMed [National Library of Medicine,Bethesda, Maryland], EMBASE [Elsevier, Amsterdam, The Netherlands];start of records until June 30, 2012) and manual bibliography searchto identify all available studies of HC in children with ASD. Electronicliterature search terms included: AUT*, ASPERGER*, ASD, PERVASIVE

www.sobp.org/journal

DEVELOPMENTAL DISORDER, PDD, HEAD CIRCUMFERENCE, OFC, andOVERGROWTH. For a study to enter our initial pool, it had to include1) children with ASD under age 5 years or an ASD group with meanage ≤10 years; and 2) a comparison of ASD HC values with those incontrol children or HC norms. Our requirements regarding study agewere designed to capture all studies including children. Of the 43studies thus identified (Figure S1 in Supplement 1), 34 survivedexclusion criteria (Table 1). Three authors (A.R., G.L.W., L.A.) independ-ently abstracted data from all 34 studies using a common set ofrules (Table S2 in Supplement 2), and any interrater inconsistencieswere resolved by consensus.

To graphically integrate reported early-life HC trajectories forchildren with ASD and control subjects, we averaged reportedmean group HC estimates across studies at standard pediatrichealth surveillance checkpoints (birth, 2, 4, 6, 9, 12, 15, and 18months). Whenever averaging HC values (e.g., HC centile ormacrocephaly rate) across studies, each reported HC value wasweighted by the number of HC observations per month uponwhich it was based. This weighting is needed to reflect the factthat 100 children aged 10 months � 2 weeks provide a morevalid estimate of mean HC at 10 months than 100 children aged10 months � 4 months.

Chi-squared tests were used to quantify the relationship betweencross-sectional study outcome (report of statistically significantevidence of HC enlargement in ASD versus report of no suchevidence) and use of HC growth norms (HC norms vs. recruitedcontrol subjects) (see text in Supplement 1 for further details). Thisapproach was adopted after testing for and ruling out the presence ofan association between ASD sample size and study outcome (p ¼ .5for macrocephaly reports, p ¼ .3 for HC centile reports).

Analysis of New HC DataParticipants. We included a total of 57 male subjects, who

were enrolled at approximately 4 years of age and comprised 22locally recruited typically developing control subjects, and 35clinically ascertained children with ASD who met DSM-IV diag-nostic criteria for autistic disorder. Full details of procedures forparticipant recruitment, screening, medical investigation, andcognitive assessment are provided in Supplement 1. In all cases,written informed consent was obtained from the participant’sparent(s). A National Institutes of Health Institutional ReviewBoard approved this study.

Head Circumference Data. Head circumference data wereretrospectively gathered for all participants from medical records.As expected, HC measures were clustered at birth and 2, 4, 6, 9,12, 15, and 18 months of age [as per American Academy ofPediatrics Recommendations for Preventative Pediatric HealthCare (42)]. Our sample included a total of 330 HC measures(201 ASD, 129 control subjects). Head circumference values wereanalyzed in both their raw form and after conversion to age- andsex-normed HC centile using CDC (34), World Health Organization(WHO) (43), and PCN (30) norms.

We note that the number of participants and overall HCmeasurements in our sample was smaller than some prior reports(Table 1) and therefore less powered to identify statisticallysignificant group differences in HC. However, in the context ofour systematic review/meta-analysis of prior data, the mainpurpose of analyzing our own raw HC data was to 1) determineif our newly derived HC trajectories would independently repli-cate our systematic review findings; 2) allow illustration ofdifferent HC norm effects on the same raw data; and 3) allowillustration of the distributional properties that arise whennormally distributed raw HC data are converted to centile.

Table 1. Summary of Studies Included in Systematic Review

Cases Comparison DataControl forHC Modifiers Findings: Coded

First Author Ref.AnalyticApproach Diagnosis Source

SampleSize

Age Range(Months)

HCObservationsper Month

HC Norm� 3 Years

HC Norm� 3 Years

ControlSubjects

HC SummaryUsed Sex Race

BodySize HC

HCChange

Controllingfor BodySize

Bolton 44 C AUT Clinic 27 60–192 .2 – Tanner N Centile cutoffs Y Y N m – –Davidovitch 45 C AUT Clinic 148 8–84 2 Nellhaus Nellhaus N Centile cutoffs Y N N m – –Woodhouse 46 C ASD Clinic 37 60–192 .2 – Tanner N Centile cutoffs Y Y N m – –Lainhart 47 C AUT Community 91 0–456 .2 Roche Roche N Mean and centile

cutoffsY N N m – –

Rodier 48 C AUT Epidemiologic 61 72–168 .7 – Feigold Y Centile cutoffs Y Y N - – –Fombonne 49 C ASD Clinic 126 14–192 .7 Sempe Sempe N Centile cutoffs Y N N m – –Fidlera 50 C ASD Clinic 41 60–192 .3 Roche Roche N Centile cutoffs Y Y N m – –Miles 51 C AUT Clinic 137 12–492 .3 Nellhaus Nellhaus N Mean and centile

cutoffsY N N m – –

Gillbergb 52 C ASD Clinic 100 0–192 44 Karlberg Knudtzon N Mean Y N N m – –Courchesnec 15 C and L ASD Clinic 48 0–14 7.5 CDC – N Mean Y N N m m –Deutsch 53 C ASD Community 63 52–168 .5 – Farkas N Mean and centile

cutoffsY Y N m – –

Torrey 54 C AUT Community 15 0–84 3.5 – – Y Mean and centilecutoffs

Y Y N - – –

Dementieva 16 C and L AUT Community 251 0–216 42 CDC Roche N Mean and centilecutoffs

Y N N m m –

Hazlett 17 C and L AUT Clinic 113 0–48 2.7 CDC – Y Mean Y Y Y - m mDissanayake 18 C and L ASD Clinic 28 0–36 .8 – – Y Mean Y N N - m –Lainhartd 15 C AUT Community 208 36–564 21.7 Roche Roche Y Mean and centile

cutoffsY Y Y m – m

Dawson 19 L ASD Community 28 0–36 5 CDC – N Mean Y N Y m m –Mills 56 C ASD Clinic 71 48–96 1.5 CDC – Y Mean Y Y Y m – mMraze 35 C ASD Community 35 0–24 1.5 CDC – Y Mean Y Y Y - – -Sacco 57 C AUT Clinic 241 36–192 1.5 NCHS Bushby N Centile cutoffs Y N N m – –van Daalen’ 58 C ASD Community 53 1–14 53 Fredriks – Y Mean Y Y Y - – -Webb 20 C and L ASD Clinic 28 0–36 8.3 CDC – N Mean Y Y N m m –Fukumoto 59 C ASD Clinic 85 0–12 42 Japanese

NormsJapaneseNorms

N Mean Y Y Y m – -

Constantino 24 L ASD Clinic 48 0–36 7.7 – – Y Mean Y N N – m –Barnard–Bark 60 C ASD Epidemiologic 100 9–36 100 WHO – Y Mean and centile

cutoffsY N N - – –

Chawarskaf 21 C and L ASD Clinic 98 0–24 98 CDC – Y Mean and centilecutoffs

Y Y Y m m -

Davidovitch 61 C ASD Clinic 317 8–84 4 CDC Nellhaus Y Mean and centilecutoffs

Y N N m – –

Fukumoto 62 C ASD Clinic 280 0–12 136 – – Y Mean Y Y Y m – mNordahl 22 C and L ASD Clinic 78 0–18 26 – – Y Mean Y Y Y m m mOzgen 63 C ASD Clinic 224 36–216 1.2 – – Y Mean Y Y Y m – m

A.Razn

ahan

etal.

BIO

LPSY

CHIATRY

2013;74:563–575

565

www.sobp.org/journal

Table 1. (continued )

Cases Comparison DataControl forHC Modifiers Findings: Coded

First Author Ref.AnalyticApproach Diagnosis Source

SampleSize

Age Range(Months)

HCObservationsper Month

HC Norm� 3 Years

HC Norm� 3 Years

ControlSubjects

HC SummaryUsed Sex Race

BodySize HC

HCChange

Controllingfor BodySize

Whitehouse 64 C ASD Community 14 0 14 Beeby – Y Mean Y Y Y - – -

Rommelse 65 C and L ASD Community 129 0–19 6.8 Fredriks – Y Mean and centilecutoffs

Y Y Y - – m

Gray 25 C and L AUT Clinic 75 0–69 37 CDC – Y Mean Y N N m - –Ververi 66 C ASD Clinic 222 18–108 2.5 Chiotis Chiotis N Centile cutoffs Y N N m - –Raznahan – C and L AUT Community 37 0–18 27 CDC,

PCN,WHO

– Y Mean and centilecutoffs

Y Y Y - - -

For the 34 studies identified, more recent studies have tended to have a tighter age range (p ¼ .04) and greater density of HC observations (p ¼ .02) (as determined by linear regression ofpublication year on age range and HC observations per month, respectively). The number of HC studies published per year has risen steeply in the past 10 years, with recent studies more often usinglongitudinal as opposed to purely cross-sectional designs. With the exception of two epidemiologically based studies (58,68) and three that identified clinical participants through broad-basedscreening efforts (23,59,69), all available studies examined clinically ascertained samples of children with ASD. In terms of clinical profile, most studies [except (18,50)] have examined groups ofparticipants with ASD unselected for general cognitive ability. While most studies have sought to exclude cases of ASD associated with potential causal genetic syndromes, the comprehensivenessof laboratory-based assays for such syndromes has been highly variable. Control for known modifiers of HC has been mixed: while all studies have accounted for sex differences in HC by use of sex-specific norms or stratification by sex, control for potential race and body size influences on HC has been much less consistent (�30% of studies simultaneously control for race and body size effectson HC). See Table S1 in Supplement 1 for more detailed study information, and see Table S2 in Supplement 1 for rules applied during data abstraction.

↑, evidence for abnormal brain enlargement (if cross-sectional finding) or accelerated brain growth (if longitudinal finding); -, no evidence found for brain enlargement or accelerated braingrowth; –, not applicable; ASD, autism spectrum disorder; AUT, autistic disorder; C, cross-sectional; CDC, Centers for Disease Control and Prevention; HC, head circumference; L, longitudinal; N, no;NCHS, National Children's Health Study; PCN, Primary Care Network; WHO, World Health Organization; Y, yes.

aInformation about age range was not available from paper or authors and is therefore taken to be 5 to 16 years for calculation of HC observations per month.bReported macrocephaly rates not abstracted as only provided relative to body size.cLongitudinal analyses compared ASD participants with a subset of individuals included in Roche HC norms.dRequired information (head size data for those under 18 years of age) only available for the autism subsample of a larger (n ¼ 338) ASD cohort.eClassified as a negative cross-sectional study since no evidence of brain enlargement relative to control subjects.fClassified as a longitudinal study based on use of repeat measures to model trajectory of HC growth, but no direct test made for differences in rate of HC change between ASD and control

subjects.

566BIO

LPSY

CHIATRY

2013;74:563–575

A.Razn

ahan

etal.

www.sobp.org/journal

Table 2. Head Circumference Reference Norms Used to Test the EBO Hypothesis in ASD

Representative HCCentiles

Age 1Year

Age 10Years

HC NormChart

Age Range(Year of Collection) Source/Ethnicity 50th 97th 50th 97th Uses in ASD Research (cf. Table 1) [Ref.]

Recent Evidence for BiasToward Overestimating

HC Centile

Tanner(1966)

0–19 years (1966) UK/Caucasian 46.7 49.6 52.7 55.2 Bolton (1994), Woodhouse (1996) [44,46] Wright (2002),Ounsted (1985)

Nellhaus(1968)

0–18 years (1943–1965) USA, MulticenterEurope/Caucasian andAfrican-American

47 49.6 53 55.9 Davidovitch (1996, 2011), Miles (2000)[45,61,51]

No studies available

NCHS(1979)

0–3 years (1930–1975)"Fels data"

USA/Caucasian 47 – – – Sacco (2007) [57] No studies available

2–18 years (1962–1974) USA/CaucasianSempe(1979)

0–19 years (1953–1954) France/Caucasian

47 49.4 53.1 55.7 Fombonne (1999) [49] No studies available

Roche(1987)

0–18 years (1930–1982)"Fels data"

USA/Caucasian 47 49.6 53.5 56.5 Lainhart (1997, 2006), Fidler (2000),Dementieva (2005) [47,55,50,16]

No studies available

Knudtzon(1988)

0–4 (1982–1984) Norway/Caucasian

46.5 49.2 53.9 55.7 Gillberg (2002) [52] No studies available

3–17 (1971–1974)

Farkas(1994)

6–18 years (1973–1976) Canada/Caucasian

– – 53.5 56.3 Deutsch (2003) [53] No studies available

CDC(2000)

0 (1960–1994) "Fels data" USA/Caucasian 46.2 48.7 – – Courchesne (2003), Dementieva (2005),Hazlett (2005), Dawson (2007), Mills (2007),Mraz (2007), Webb (2007), Chawarska(2011), Davidovitch (2011), Gray (2012)[15,16,17,19,56,35,20,21,61,25]

Karvonen (2012),Daymont (2010),Schienkiewitz (2011),Werner (2006)

2 months–3 years (1971–1974 NHANES I)

USA/NationallyRepresentative

6 months–3 years (1976–1980 NHANES II)

USA/NationallyRepresentative

2 months–3 years (1988–1994 NHANES III)

USA/NationallyRepresentative

Fredriks(2000)

2 weeks–21 years (1996–1997)

The Netherlands/Caucasian

47.2 50.2 53.3 56.7 van Daalen (2007), Rommelse (2011) [58,65] No studies available

WHO(2006)

0–5 years (1997–2003) Global/Multiethnic

46.1 48.5 – – Barnard-Brak (2011) [60] Daymont (2010),Juliusson (2011)

Details for HC norms used in HC tests of the EBO hypothesis. Three norm sources are not included, as the required information could not be obtained[Feingold and Bossert (78); Chiotis et al. (79); Education (80)].

ASD, autism spectrum disorder; CDC, Centers for Disease Control and Prevention; EBO, early brain overgrowth; HC, head circumference; NCHS, NationalChildren's Health Study; NHANES, National Health and Nutrition Examination Survey; UK, United Kingdom; USA, United States of America; WHO, WorldHealth Organization.

A. Raznahan et al. BIOL PSYCHIATRY 2013;74:563–575 567

StatisticsAfter assessing if HC indices at each age point violated the

assumption of normality (Shapiro-Wilk test), parametric (Studentt) or nonparametric (Wilcox signed rank) tests were applied asappropriate to carry out serial cross-sectional comparisons of 1)children with ASD versus control subjects for both raw HC and HCcentile, and 2) HC centile in children with ASD and controlsubjects versus the norm-predicted HC centile value of 50.

Longitudinal analysis of HC data was first conducted bycomparing changes in HC between adjacent age points usingthe Mann-Whitney U tests (given nonnormality of HC changeaccording to Shapiro-Wilks test). Next, for the 44 participants (26ASD, 18 control subjects) with at least five HC measures, we usednonlinear mixed models to model HC growth according to thefollowing asymptotic regression equation:

HC ¼ f1�f2½eðf3*tÞ�

This growth curve was chosen for consistency with priornonlinear modeling of HC in ASD (17,24). Sequential model fittingwith F test comparison between models was used to specifyrandom effects and determine if ASD and control groups showedsignificant differences in HC growth (either for individual param-eters or across all parameters).

Results

Systematic ReviewTable 1 lists the 34 studies identified by our systematic review

(15–22,24,25,35,44–66), according to publication date and studycharacteristics.

Survey of Normative Data Used to Test EBO Hypothesis inASD. Of the 34 studies identified, 21 studies made exclusive useof HC norms as comparison data when testing the EBO hypoth-esis, 8 studies used both norms and locally recruited control

www.sobp.org/journal

568 BIOL PSYCHIATRY 2013;74:563–575 A. Raznahan et al.

subjects, and 5 studies made exclusive use of locally recruitedcontrol subjects. Head circumference norms used in tests of theEBO hypothesis vary greatly in the historical, geographical, andethnic properties of the underlying samples they are based upon(Table 2). While cross-sectional HC studies in ASD have drawn ona diverse pool of HC norms, 88% of all longitudinal studies haverelied on CDC norms.

Cross-Sectional Studies. Thirty-two (94%) of the 34 studiesidentified by systematic review included a cross-sectionalcomparison of HC between participants with ASD and previ-ously published norms or HC in locally recruited controlsubjects. Cross-sectional comparisons between HC in childrenwith ASD and locally recruited control subjects more oftenfailed to find evidence for EBO than when comparisons werebased on previously published HC norms (χ2 ¼ 10.8, p ¼ .001)(Figure 1A).

Control for body size has had mixed effects on study outcome:five studies found that significant HC abnormalities inASD remained even when body size was included as a covariate(22,55,56,62,63), but five studies failed to find significantHC alterations in ASD after covarying for body size (21,35,58,59,64).

Cross-sectional HC studies provide two types of summarymetrics amenable to integration across multiple reports: macro-cephaly rate and mean HC centile. These integrative analyses arefully detailed in the text and Figure S2 in Supplement 1 and TableS3 in Supplement 2. For the more commonly used index ofmacrocephaly rate (Figure 1B), we found that 1) exclusion of the38% of reports that estimated macrocephaly rates in ASD usingHC norms that are now known to overidentify brain enlargementin contemporary samples of healthy children (Table 2) reducedthe mean ASD macrocephaly rate across all studies from 14% to4%; and 2) studies that compared macrocephaly rates in ASDagainst those in contemporaneously ascertained control subjectswere significantly less likely to identify significantly elevatedmacrocephaly in ASD when compared with studies comparingobserved rates in ASD with the norm-predicted rate of 3%(χ2 ¼ 10.5, p ¼ .002). A parallel effect was seen for HC centilefindings: similar group mean HC centile estimates in ASD tend tohave been differently interpreted based on whether expected(i.e., 50th HC centile) or observed (i.e., HC centile in recruitedcontrol sample) HC centile data were used as a comparison(Figure S2 in Supplement 1).

Longitudinal Studies. Eleven (32%) of the 34 studies identi-fied by our systematic review included a longitudinal comparisonof HC in a group of participants with ASD with HC norms and/orcontrol subjects (Figure 1C). The four norm-referenced longitudi-nal HC studies in ASD have all used CDC HC norms(15,16,19,20,67) and consistently found evidence for robust EBOin ASD during the first year of life. In contrast, reports based oncomparisons with locally recruited control subjects have beenmixed. The two control group referenced studies that imposelinear HC trajectories within selected age bands (22,24) replicatereports of EBO in ASD versus CDC norms, i.e., abnormallyaccelerated HC increase during the first year of life. Other controlgroup referenced studies either find evidence of subtle EBO inASD that emerges during the second year of life (17,18) or fail tofind any evidence for abnormally accelerated HC growth in ASD(23,25).

Integrating Cross-Sectional and Longitudinal HC DataBetween Birth and 18 Months. Of the 16 studies that provideHC data within the first 2 years of life, 4 provide sex-normed CDCHC centiles (15,19,20,35) and 5 others provide sex-specific

www.sobp.org/journal

estimates of mean raw HC that can be converted into HC centile(17,18,21,22,25). Taken together, these nine studies provideinformation about HC growth in 470 children with ASD and 342control subjects, using a total of 2005 and 1814 HC observations,respectively (Table S4 in Supplement 2). The deviant CDC HCcentile trajectories that these studies all identify in ASD aremimicked by PCN norms—the largest and most recent descriptionof HC growth in US children (30) (Figure 2A). This same pattern ofHC overgrowth relative to CDC norms is also clearly evident in atleast three (18,22,35) of the four (18,21,22,35) samples of typicallydeveloping children that have been included as control subjects inASD studies (Figure 2B). Figure 2C shows cross-study average CDCcentile trajectories of HC growth for clinically ascertained childrenwith ASD and control groups used in ASD studies, alongside the50th centile for HC growth in PCN norms.

Analysis of New HC DataParticipant characteristics are summarized in Table 3.Serial Cross-Sectional Analyses. Distributions for raw HC

did not show statistically significant violations from normalityfor either group at any age point. There were no statisticallysignificant group differences in raw HC at any age point(Table 4). Head circumference centile distributions were signifi-cantly nonnormal for both ASD and control groups at most agepoints examined, although these deviations were more marked forCDC and WHO than PCN norms (see text and Figure S3 inSupplement 1 for more details). Nevertheless, to allow comparisonwith prior HC studies in ASD (which commonly report mean HCcentile values at each age point [Table S1 in Supplement 2]), weplot mean cross-sectional HC centile values for CDC, WHO, andPCN in Figure 3. For HC in both the ASD and control group 1)mean CDC centile rises steeply from birth to reach a plateau atabout 8 months; 2) mean WHO centile is within expected limits atbirth, but abnormally elevated thereafter; and 3) mean PCN centileadheres most closely to the expected flat trajectory (Figure 3C).

Longitudinal Analyses. There were no statistically significantgroup differences in rate of HC change between any two HCmeasurement time points after correction for multiple compar-isons (Table 4). Nonlinear mixed modeling of HC growth(Figure 4A) did not identify any significant trajectory differencesbetween ASD and control groups, although visual comparisonof the best fit trajectory for each group suggests a steadydivergence of mean HC such that by 18 months, estimated meanHC in ASD is �7 mm greater than that in control subjects(equivalent to an effect size of .5 d based on the pooledstandard deviation of HC across ASD and control groups at age18 months).

Our findings in cross-sectional and longitudinal HC analyseswere robust to stratification by developmental regression statusand control for height and demographic variables (see text inSupplement 1).

Integration With Prior HC Data Between Birth and 18Months. Figure 4B shows that HC growth for the newly presentedsample of children with ASD and control subjects replicates thepattern of HC overgrowth in the first year of life relative to CDCnorms that has been found in previous studies of children withASD and control subjects, as well as the largest and most recentlypublished set of US HC norms (30). This highly replicable patternof EBO during the first year of life is abolished when HC values arere-expressed as their equivalent male PCN centile (Figure 4C):patterns of HC growth in previous studies and in our newlypresented cohort are broadly concordant with those predicted byPCN norms.

Figure 1. Systematic review findings. (A) Cross-sectional studies (in reverse order of publication year) grouped according to use of head circumference (HC) normsversus locally recruited control subjects as a comparison for autism spectrum disorder (ASD) data: this grouping was significantly associated with study outcome (χ2¼ 10.8, p ¼ .001). A negative study outcome was also associated with a narrower study age range (t ¼ 2.3, p ¼ .03). Each study is a horizontal bar spanning thestudy age range, with study outcome coded by color and density of HC observations coded by bar height. Where one study reported multiple cross-sectional tests,the coded outcome relates to analysis of raw HC where available and HC centile where not available. Outcomes are coded before any control for body size. (B)Summary of reported macrocephaly rates in ASD grouped according to use of expected macrocephaly rate in HC norms (3%) or observed macrocephaly rate incontrol subjects as a comparison: each reported rate is a point, with significance (SIG') of test for macrocephaly enrichment in ASD coded by shape; HC norm iscoded by color and density of HC observations is coded by size. Study stratification by comparison group was significantly associated with study outcome (χ2 ¼10.5, p¼ .005). (C) Longitudinal studies grouped according to use of HC norms versus locally recruited control subjects as a comparison for ASD data: each study is ahorizontal bar spanning the study age range, with periods of abnormal HC change in ASD coded by color and density of HC observations coded by bar height. CDC,Centers for Disease Control and Prevention; NCHS, National Childrens Health Study; NCPP, National Collaborative Perinatal Project; WHO, World Health Organization.

A. Raznahan et al. BIOL PSYCHIATRY 2013;74:563–575 569

www.sobp.org/journal

Figure 2. Reported patterns of head circumference (HC) growth betweenbirth and 18 months within the common framework of Centers for DiseaseControl and Prevention (CDC) head circumference reference norms. Thetrajectory for expected growth according to CDC norms (50th centile) isshown as a horizontal solid black line. The trajectories for expected growthaccording to Primary Care Network (PCN) norms (50th centile) is shown asdashed black line. (A) Prior reports in autism spectrum disorder (ASD): eachstudy is a differently colored solid line. (B) Prior reports for control groupsincluded in studies of HC in ASD: each study is a differently colored solidline. (C) Summary of weighted mean ASD (solid red) and control (solid blue)HC trajectories. Confidence intervals are not shown as constituent studiesdid not uniformly provide estimates of error for reported group HC means.

570 BIOL PSYCHIATRY 2013;74:563–575 A. Raznahan et al.

www.sobp.org/journal

Discussion

Our systematic review identifies several lines of evidence thatthe use of HC population norms has introduced systematic age-dependent biases in studies of early brain growth in ASD.We were able to replicate and better specify the nature of theseage-dependent HC norm biases by analyzing a new set of HCdata in children with ASD against different sources of controldata.

The Influence of HC Norm Use on Tests of the EBO HypothesisThe vast majority (83%) of HC studies testing the EBO

hypothesis in ASD involve the use of HC norms as comparisondata. At first glance, these studies provide strong support forthe hypothesis that children with ASD have an abnormalacceleration of brain growth in the first year of life(15,19,20,25,35) and that it persists as an HC excess once theearly childhood years of rapid HC growth have passed(16,44,46,50,51,53,55,57). However, several findings in our study,when considered alongside existing knowledge regarding secularchanges in HC growth [United Kingdom (28,36), China (37),Netherlands (38), Finland (32), Japan (39), and Korea (40)]and the improper approximation of contemporary HC growthpatterns by commonly used HC norms (28–33), suggest thatthe striking consistency of norm-based findings in ASD mayreflect replicable biases in HC norms rather than a replicablepattern of dysregulated growth in ASD.

First, we show that comparisons of HC data between childrenwith ASD and locally recruited control subjects are significantlyless likely to identify HC abnormalities in ASD than are studiesusing HC norms. Second, removal of those cross-sectional reportsthat use HC norms with known biases toward the overidentifica-tion of HC enlargement reduces the estimated mean macro-cephaly rate in ASD from 14% to 4%. Third, and critical to theappraisal of evidence for EBO in ASD, we find that the reporteddevelopmental unfolding of EBO in ASD relative to CDC norms[(15,19,20,25,35) and the present study)—the most commonlyused HC norm in ASD research—is closely mimicked by CDC HCcentile growth in 1) control samples included in prior ASD studies(Figure 2B and Figure S3 in Supplement 1); 2) control subjects inour own study (Figure 3); and 3) �500k HC observations in themost recently published set of HC norms for US children Figures 2and 4 and (30)]. Thus, systematic review, and the new dataanalyzed here all find that HC growth in healthy children startsslightly below what is expected by CDC norms, crosses the 50thCDC centile at between 3 and 5 months of age, and continues tocross CDC centiles until stabilizing at approximately 8 to 10months. These convergences indicate that prominent reports ofsuch HC growth in children with ASD (15,19,20,25,35), whichaccount for a large degree of the replicability across studies thattest for EBO in the first 2 years of life, reflect replicable CDC normbiases rather than a disease-related phenomenon.

The evidence we present for norm-related biases has con-sequences that stretch far beyond research on aberrant earlybrain growth in neurodevelopmental psychiatry. Comparisonswith population growth norms are an integral part of everydayclinical practice and are also frequently referred to in publichealth settings (68). In both these contexts, decisions informed bygrowth norm use can carry significant cost and risk. Our studyprovides a strong and empirically grounded argument for use ofmultiple growth norms in parallel, as well as comparison withdata from typically developing and clinical control samples [e.g.,(23)] where possible.

Table 3. Participant Characteristics

Characteristic Control ASD Group Diffference

Number 22 35ADOS Severity Score, Mean (SD) 1 (.4) 7 (4.1) p � .0005Full DQ, Mean (SD) 113 (11.3) 53 (18.4) p � .0005Verbal DQ, Mean (SD) 111 (17.9) 45 (22) p � .0005Nonverbal DQ, Mean (SD) 115 (12.3) 61 (17.2) p � .0005SES, Mean (SD) 59 (3.9) 56 (7.1) p ¼ .07Age at HC Measurement (Months), Mean (SD)

Birth (n control Subjects ¼ 18 / n ASD ¼ 26) .01 (.02) .04 (.07) ns2 Month Age Point (18/27) 1.95 (.15) 2.04 (.16) ns4 Month Age Point (16/27) 4.04 (.1) 4.03 (.13) ns6 Month Age Point (18/24) 6.03 (.16) 6.07 (.16) ns9 Month Age Point (17/25) 9.01 (.13) 9.07 (.14) ns12 Month Age Point (13/22) 12.08 (.09) 12.15 (.15) ns15 Month Age Point (16/26) 15.08 (.52) 15.23 (.35) ns18 Month Age Point (13/24) 18.20 (.53) 18.34 (.42) ns

Total Number of HC Measures 129 201Mean Number of HC Measures Per Person 5.9 5.7

ADOS, Autism Diagnostic Observation Schedule; ASD, autism spectrum disorder; DQ, developmental quotient; HC, head circumference; ns,nonsignificant; SD, standard deviation; SES socioeconomic status [as estimated by Hollingshead (81)].

A. Raznahan et al. BIOL PSYCHIATRY 2013;74:563–575 571

Evidence for EBO in ASD from HC Studies that MinimizeNorm-Related Biases

Thirteen studies—six cross-sectional (54,56,58,60,62,63) andseven longitudinal (17,18,21–25)—have employed study designsthat minimize the potential for norm-related biases when testingthe EBO hypothesis. The bias-minimizing strategies are: 1) directcomparisons of raw HC in ASD with that in control subjects(17,18,21–25,56,60,62,63); 2) epidemiologic sampling of childrenwith ASD and determination of HC centile relative to healthycontrol subjects identified within the same sampling frame (54);and 3) selection of published HC norms that are secularly anddemographically close to the ASD sample in question (23,58).

Of the six cross-sectional studies that minimize norm-relatedbiases, three find evidence of HC enlargement in ASD (56,62,63)and three do not (54,58,60). The three negative reports alldraw on population or broad screening-based samples ofchildren with ASD. The three positive cross-sectional studiesall examine clinically ascertained samples and identify mild

Table 4. Serial Analysis of HC at Each Age Point and HC Change Between Ea

Head Circumference

Control Subjects ASD

Age (Months) n Mean (SD) n Mean (SD) p Age

0 18 34.6 (1.23) 26 34.4 (1.26) .2 –2 18 39.4 (1.31) 27 39.8 (2.10) .4 0–24 16 42.4 (1.0) 27 42.5 (1.24) .7 2–46 18 44.1 (1.04) 24 44.7 (1.53) .05 4–69 17 45.9 (1.20) 25 46.1 (1.47) .5 6–912 13 46.7 (1.01) 22 47.4 (1.44) .1 9–115 16 47.7 (1.04) 26 48.2 (1.6) .2 12–118 13 48.5 (.9) 24 49.1 (1.5) .2 15–1

The borderline significant p ¼ .05 6 mm excess in mean ASD HC relative tocomparisons (corrected p value ¼ .4). Group differences in raw HC2 and 4 months of age when the HC increase in control subjects was greaterdid not survive correction for multiple comparisons (corrected p value ¼ .06). Fo(12 control subjects and 19 ASD), total HC change between birth and 18 mogroups (mean [SD] HC increase in cm: control subjects –14.2 [1.1], ASD – 14.3

ASD, autism spectrum disorder; HC, head circumference; SD, standard dev

(4–7 mm) degrees of HC enlargement among male subjectswith ASD [two after stratification by sex (62,63) and one byvirtue of only including male subjects (56)]. The reported effectsize for these HC excess in ASD is � .5 d, which converges withthat found in our sample, and is approximately half that identifiedin reports of mean HC excess in ASD relative to HC norms(Figure 2).

Of the seven longitudinal studies that minimize norm-relatedbiases, four (17,18,22,24) identify periods in early childhood whenthe rate of HC increase in clinically ascertained children with ASDis faster than that in control subjects. Two (22,24) of these fourpositive studies force HC growth to be linear between selectedbreakpoints and identify an early phase of abnormally acceleratedHC growth in ASD that coincides with that arising through bias inCDC norms (15,19,20,25,35). In contrast, the two positive longi-tudinal studies that model HC growth using more biologicallyvalid exponential growth curves find that the group average ASDgrowth curves are essentially identical to those in control subjects

ch Pair of Age Points

Head Circumference Change

Control Subjects ASD

Difference n Median (Range) n Median (Range) p

– – – – –15 5 (3–7) 20 6 (2–8) .316 3 (2–5) 25 2 (1–5) .0115 2 (0–3) 24 2 (0–4) .0616 2 (1–3) 21 2 (0–3) .7

2 12 1 (0–2) 18 1 (1–3) .65 11 1 (0–1) 20 .5 (0–2) .98 11 1 (0–2) 21 1 (0–2) .2

control subjects at age 6 months did not survive correction for multiplechange approached statistical significance for the interval betweenthan that for children with ASD (p ¼ .01). This isolated group differencer the subset of participants with complete HC data at birth and 18 monthsnths was normally distributed and statistically indistinguishable between[1.4], t test for difference p ¼ .83).iation.

www.sobp.org/journal

Figure 3. Estimates of mean head circumference (HC) centile (with 95% confidence intervals) for autism spectrum disorder (ASD) and control participantsat each age point according to different HC norms. (A) Centers for Disease Control and Prevention (CDC), (B) World Health Organization (WHO), and (C)Primary Care Network (PCN) HC norms. For HC in both the ASD and control group, 1) mean CDC centile rises steeply from birth to reach a plateau at about8 months, 2) mean WHO centile is within expected limits at birth but abnormally elevated thereafter; and 3) mean PCN centile adheres most closely to theexpected flat trajectory.

572 BIOL PSYCHIATRY 2013;74:563–575 A. Raznahan et al.

until the second year of life (17,18). These four positive studiesstand alongside three negative longitudinal reports—one ofwhich finds that growing differences between HC in ASD andtypically developing control subjects are in keeping with moregeneralized growth abnormalities (21) and another two that findno evidence of accelerated HC growth in children with ASDrelative to control subjects (23,25).

Thus, to the extent that existing (albeit nonepidemiologic)data support the presence of a generalized tendency toward EBOin ASD, this overgrowth appears to be much more subtle andlater onset in nature than the picture of gross overgrowth duringinfancy that emerges due to norm biases. These distinctions arebiologically relevant, given that regional contributions to globalbrain volume change vary so markedly between the first andsecond year of life (69).

Figure 4. Growth curve modeling of head circumference (HC) data in newsystematic review. (A) Patterns of HC change in new cohort of participants wdepicted as a set of connected points, with overall group best-fit trajectories shblue). (B) Group best-fit longitudinal HC trajectories in new cohort and weightefor Disease Control and Prevention (CDC) centile. The trajectories for expected gblack line. The trajectories for expected growth (50th centile) according to Primbest-fit longitudinal HC trajectories in new cohort and weighted mean HCtrajectories for expected growth (50th centile) according to PCN norms is showconstituent studies did not uniformly provide estimates of error for reported

www.sobp.org/journal

Summary and Next StepsIn conclusion, the most methodologically robust and bias-free

sources of evidence regarding HC growth in ASD—longitudinalstudies that model and compare HC growth curves in ASD withthose in carefully matched control subjects (17,18,21,23,25)—areequivocal regarding the presence of abnormally accelerated earlybrain growth in ASD. Where evidence for EBO has been found, HCgrowth in ASD begins to show a subtle divergence from that incontrol subjects during the second year of life (17,18). Thispattern of EBO is distinct from the dramatic brain overgrowthbefore age 1 year that is suggested by CDC norm analyses(15,19,20,25,35) and its presence has yet to be established inpopulation-based samples of children with ASD. This latterlimitation is of particular concern given that well-establisheddemographic differences between children with ASD who are

cohort and combination of new cohort findings with Summaries fromith autism spectrum disorder (ASD) and control subjects. Each person isown as bold solid lines: color-coded for group (ASD: red, control subjects:d mean HC trajectories generated by systematic review, shown as Centersrowth (50th centile) according to CDC norms is shown as a horizontal solidary Care Network (PCN) norms are shown as dashed black line. (C) Grouptrajectories generated by systematic review, shown as PCN centile. Then as a horizontal solid black line. Confidence intervals are not shown, as

group HC means.

A. Raznahan et al. BIOL PSYCHIATRY 2013;74:563–575 573

known to clinical services versus those that remain undiagnosed[such as maternal educational level (70)] are also highly correlatedwith HC growth in early childhood (71). Such confounds meanthat recruitment biases in nonepidemiologic clinical samples maylead to artifactual case-control differences in HC growth. Withappropriate consideration of these methodological limitations,studies of HC may still provide a valuable tool for biologicalresearch in ASD, however, by virtue of being much moreamenable than brain structural magnetic resonance imaging tolongitudinal research in large population-based samples.

Moving forward, it will be critical to determine if epidemio-logically ascertained samples of children with ASD show longi-tudinal disruptions of HC growth relative to norms defined withinthe same population-based reference frame or contemporaneouslocal control subjects (rather than historical norms). Analyses ofHC growth in ASD within more recently established population-based cohorts (72) could address these limitations, however, andalso help to clarify if aberrant brain growth in ASD segregateswith characteristics such as sex and regression status (22).

While a generalized tendency toward HC increase in ASDremains to be established, there is no doubt that brain enlarge-ment is a consistent feature of several genetic syndromes that areaccompanied by an ASD-like behavioral phenotype (73). Of note,however, the opposite is also true, as genetic studies have alsoshown that ASD-like behaviors are a clinical feature of severalmicrocephalic syndromes (74). One hypothesis, therefore, is thatextreme dysregulation of brain growth may be more relevant toour understanding of ASD than brain enlargement per se. Unlikeexistent applications of HC norms to clinically ascertainedsamples of children with ASD (which have largely failed to findsignificantly elevated microcephaly rates), forthcomingpopulation-based studies with rich genomic and environmentalinformation (75) should be able to compare the mutuallyexclusive extreme growth versus isolated overgrowth hypothesesand link instances of aberrant brain size to candidate genetic/environmental influences during prenatal or early postnatal life.

Beyond clarifying the presence and nature of EBO in ASD andthe extent of any accompanying association between ASD andmicrocephalic syndromes, future research will ultimately need todetermine where disruptions of overall brain growth in ASD sit onthe spectrum between epiphenomenal biomarker to causalintermediate phenotype (76). The answer to this question—whichwill rely heavily on longitudinal and genetically informativestudies in developing humans [e.g., Infant Brain Imaging Study(77)]—ultimately determines the potential for continued study ofEBO to improve etiological understanding and clinical manage-ment of ASD.

Financial support for this work was provided by the NationalInstitutes of Health Intramural Research Program.

We thank the children and families who took part in this study.All authors report no biomedical financial interests or potential

conflicts of interest.

Supplementary material cited in this article is available online athttp://dx.doi.org/10.1016/j.biopsych.2013.03.022.

1. Courchesne E, Karns CM, Davis HR, Ziccardi R, Carper RA, Tigue ZD,et al. (2001): Unusual brain growth patterns in early life in patientswith autistic disorder: An MRI study. Neurology 57:245–254.

2. Courchesne E, Campbell K, Solso S (2011): Brain growth across the lifespan in autism: Age-specific changes in anatomical pathology. BrainRes 1380:138–145.

3. Veenstra-VanderWeele J, Blakely RD (2012): Networking in autism:Leveraging genetic, biomarker and model system findings in thesearch for new treatments. Neuropsychopharmacology 37:196–212.

4. Page DT, Kuti OJ, Prestia C, Sur M (2009): Haploinsufficiency for Ptenand Serotonin transporter cooperatively influences brain size andsocial behavior. Proc Natl Acad Sci U S A 106:1989–1994.

5. Foldi CJ, Eyles DW, McGrath JJ, Burne TH (2010): Advanced paternalage is associated with alterations in discrete behavioural domains andcortical neuroanatomy of C57BL/6J mice. Eur J Neurosci 31:556–564.

6. Chow ML, Pramparo T, Winn ME, Barnes CC, Li HR, Weiss L, et al.(2012): Age-dependent brain gene expression and copy numberanomalies in autism suggest distinct pathological processes at youngversus mature ages. PLoS Genet 8:e1002592.

7. Courchesne E, Mouton PR, Calhoun ME, Semendeferi K, Ahrens-BarbeauC, Hallet MJ, et al. (2011): Neuron number and size in prefrontal cortexof children with autism. JAMA 306:2001–2010.

8. Shinawi M, Liu P, Kang SH, Shen J, Belmont JW, Scott DA, et al. (2010):Recurrent reciprocal 16p11.2 rearrangements associated with globaldevelopmental delay, behavioural problems, dysmorphism, epilepsy,and abnormal head size. J Med Genet 47:332–341.

9. Muller RA, Shih P, Keehn B, Deyoe JR, Leyden KM, Shukla DK (2011):Underconnected, but how? A survey of functional connectivity MRIstudies in autism spectrum disorders. Cereb Cortex 21:2233–2243.

10. Lewis JD, Theilmann RJ, Sereno MI, Townsend J (2009): The relationbetween connection length and degree of connectivity in youngadults: A DTI analysis. Cereb Cortex 19:554–562.

11. Murias M, Webb SJ, Greenson J, Dawson G (2007): Resting statecortical connectivity reflected in EEG coherence in individuals withautism. Biol Psychiatry 62:270–273.

12. 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.

13. Kraemer HC, Yesavage JA, Taylor JL, Kupfer D (2000): How can welearn about developmental processes from cross-sectional studies, orcan we? Am J Psychiatry 157:163–171.

14. Bartholomeusz HH, Courchesne E, Karns CM (2002): Relationshipbetween head circumference and brain volume in healthy normaltoddlers, children, and adults. Neuropediatrics 33:239–241.

15. Courchesne E, Carper R, Akshoomoff N (2003): Evidence of brainovergrowth in the first year of life in autism. JAMA 290:337–344.

16. Dementieva YA, Vance DD, Donnelly SL, Elston LA, Wolpert CM, RavanSA, et al. (2005): Accelerated head growth in early development ofindividuals with autism. Pediatr Neurol 32:102–108.

17. Hazlett HC, Poe M, Gerig G, Smith RG, Provenzale J, Ross A, et al. (2005):Magnetic resonance imaging and head circumference study of brain sizein autism: Birth through age 2 years. Arch Gen Psychiatry 62:1366–1376.

18. Dissanayake C, Bui QM, Huggins R, Loesch DZ (2006): Growth in statureand head circumference in high-functioning autism and Aspergerdisorder during the first 3 years of life. Dev Psychopathol 18:381–393.

19. Dawson G, Munson J, Webb SJ, Nalty T, Abbott R, Toth K (2007): Rateof head growth decelerates and symptoms worsen in the second yearof life in autism. Biol Psychiatry 61:458–464.

20. Webb SJ, Nalty T, Munson J, Brock C, Abbott R, Dawson G (2007): Rateof head circumference growth as a function of autism diagnosis andhistory of autistic regression. J Child Neurol 22:1182–1190.

21. Chawarska K, Campbell D, Chen L, Shic F, Klin A, Chang J (2011): Earlygeneralized overgrowth in boys with autism. Arch Gen Psychiatry 68:1021–1031.

22. Nordahl CW, Lange N, Li DD, Barnett LA, Lee A, Buonocore MH, et al.(2011): Brain enlargement is associated with regression in preschool-age boys with autism spectrum disorders. Proc Natl Acad Sci U S A 108:20195–20200.

23. Rommelse NN, Peters CT, Oosterling IJ, Visser JC, Bons D, van Steijn DJ, et al.(2010): A pilot study of abnormal growth in autism spectrum disorders andother childhood psychiatric disorders. J Autism Dev Disord 41:44–54.

24. Constantino JN, Majmudar P, Bottini A, Arvin M, Virkud Y, Simons P,Spitznagel E (2010): Infant head growth in male siblings of children withand without autism spectrum disorders. J Neurodev Disord 2:39–46.

25. Gray KM, Taffe J, Sweeney DJ, Forster S, Tonge BJ (2012): Could headcircumference be used to screen for autism in young males withdevelopmental delay? J Paediatr Child Health 48:329–334.

26. Schumann CM, Bloss CS, Barnes CC, Wideman GM, Carper RA,Akshoomoff N, et al. (2010): Longitudinal magnetic resonance imaging

www.sobp.org/journal

574 BIOL PSYCHIATRY 2013;74:563–575 A. Raznahan et al.

study of cortical development through early childhood in autism.J Neurosci 30:4419–4427.

27. Hazlett HC, Poe MD, Gerig G, Styner M, Chappell C, Smith RG, et al.(2011): Early brain overgrowth in autism associated with an increase incortical surface area before age 2 years. Arch Gen Psychiatry 68:467–476.

28. Ounsted M, Moar VA, Scott A (1985): Head circumference chartsupdated. Arch Dis Child 60:936–939.

29. Wright CM, Inskip HM, Godfrey K, Williams AF, Ong KK (2011):Monitoring head size and growth using the new UK-WHO growthstandard. Arch Dis Child 96:386–388.

30. Daymont C, Hwang WT, Feudtner C, Rubin D (2010): Head-circumference distribution in a large primary care network differsfrom CDC and WHO curves. Pediatrics 126:e836–e842.

31. Schienkiewitz A, Schaffrath Rosario A, Dortschy R, Ellert U, NeuhauserH (2011): German head circumference references for infants, childrenand adolescents in comparison with currently used national andinternational references. Acta Paediatr 100:e28–e33.

32. Karvonen M, Hannila ML, Saari A, Dunkel L (2012): New Finnish referencefor head circumference from birth to 7 years. Ann Med 44:369–374.

33. Werner B, Bodin L (2006): Head circumference from birth to age 48months for infants in Sweden. Acta Paediatr 95:1601–1607.

34. Kuczmarski RJ, Ogden CL, Guo SS, Grummer-Strawn LM, Flegal KM,Mei Z, et al. (2002): 2000 CDC Growth Charts for the United States:Methods and development. Vital Health Stat 11:1–190.

35. Mraz KD, Green J, Dumont-Mathieu T, Makin S, Fein D (2007):Correlates of head circumference growth in infants later diagnosedwith autism spectrum disorders. J Child Neurol 22:700–713.

36. Wright CM, Booth IW, Buckler JMH, Cameron N, Cole TJ, Healy MJR,et al. (2002): Growth reference charts for use in the United Kingdom.Arch Dis Child 86:11–14.

37. Davies DP, Leung SF, Lau SP (1986): Secular trends in head growth.Arch Dis Child 61:623–624.

38. Fredriks AM, van Buuren S, Burgmeijer RJ, Meulmeester JF, Beuker RJ,Brugman E, et al. (2000): Continuing positive secular growth change inThe Netherlands 1955-1997. Pediatr Res 47:316–323.

39. Anzo M, Takahashi T, Sato S, Matsuo N (2002): The cross-sectionalhead circumference growth curves for Japanese from birth to 18 yearsof age: The 1990 and 1992-1994 national survey data. Ann Hum Biol29:373–388.

40. Moon JS (2011): Secular trends of body sizes in Korean children andadolescents: From 1965 to 2010. Korean J Pediatr 54:436–442.

41. Gozzi M, Nielson DM, Lenroot RK, Ostuni JL, Luckenbaugh DA, Thurm AE,et al. (2012): A magnetization transfer imaging study of corpus callosummyelination in young children with autism. Biol Psychiatry 72:215–220.

42. Hagan JF, Shaw JS, Duncan PM (2008): Bright Futures: Guidelines forHealth Supervision of Infants, Children, and Adolescents, 3rd ed. ElkGrove Village, IL: American Academy of Pediatrics.

43. World Health Organization Multicentre Growth Reference StudyGroup (2007): WHO Child Growth Standards: Head Circumference-For-Age, Arm Circumference-For-Age, Triceps Skinfold-For-Age and Subscap-ular Skinfold-For-Age. Methods and Development. Geneva: World HealthOrganization Press.

44. Bolton P, Macdonald H, Pickles A, Rios P (1994): A case-control familyhistory study of autism. J Child Psychol Psychiatry 35:877–900.

45. Davidovitch M, Patterson B, Gartside P (1996): Head circumferencemeasurements in children with autism. J Child Neurol 11:389–393.

46. Woodhouse W, Bailey A, Rutter M, Bolton P, Baird G, Le Couteur A(1996): Head circumference in autism and other pervasive devel-opmental disorders. J Child Psychol Psychiatry 37:665–671.

47. Lainhart JE, Piven J, Wzorek M, Landa R, Santangelo SL, Coon H,Folstein SE (1997): Macrocephaly in children and adults with autism.J Am Acad Child Adolesc Psychiatry 36:282–290.

48. Rodier PM, Bryson SE, Welch JP (1997): Minor malformations and physicalmeasurements in autism: Data from Nova Scotia. Teratology 55:319–325.

49. Fombonne E, Roge B, Claverie J, Courty S, Fremolle J (1999): Micro-cephaly and macrocephaly in autism. J Autism Dev Disord 29:113–119.

50. Fidler DJ, Bailey JN, Smalley SL (2000): Macrocephaly in autism andother pervasive developmental disorders. Dev Med Child Neurol 42:737–740.

51. Miles JH, Hadden LL, Takahashi TN, Hillman RE (2000): Head circum-ference is an independent clinical finding associated with autism. AmJ Med Genet 95:339–350.

www.sobp.org/journal

52. Gillberg C, de Souza L (2002): Head circumference in autism, Aspergersyndrome, and ADHD: A comparative study. Dev Med Child Neurol 44:296–300.

53. Deutsch CK, Joseph RM (2003): Brief report: Cognitive correlates ofenlarged head circumference in children with autism. J Autism DevDisord 33:209–215.

54. Torrey EF, Dhavale D, Lawlor JP, Yolken RH (2004): Autism and headcircumference in the first year of life. Biol Psychiatry 56:892–894.

55. Lainhart JE, Bigler ED, Bocian M, Coon H, Dinh E, Dawson G, et al.(2006): Head circumference and height in autism: A study by theCollaborative Program of Excellence in Autism. Am J Med Genet A 140:2257–2274.

56. Mills JL, Hediger ML, Molloy CA, Chrousos GP, Manning-Courtney P,Yu KF, et al. (2007): Elevated levels of growth-related hormones inautism and autism spectrum disorder. Clin Endocrinol (Oxf) 67:230–237.

57. Sacco R, Militerni R, Frolli A, Bravaccio C, Gritti A, Elia M, et al. (2007):Clinical, morphological, and biochemical correlates of head circum-ference in autism. Biol Psychiatry 62:1038–1047.

58. van Daalen E, Swinkels SHN, Dietz C, Van Engeland H, Buitelaar JK(2007): Body length and head growth in the first year of life in autism.Pediatr Neurol 37:324–330.

59. Fukumoto A, Hashimoto T, Ito H, Nishimura M, Tsuda Y, Miyazaki M,et al. (2008): Growth of head circumference in autistic infants duringthe first year of life. J Autism Dev Disord 38:411–418.

60. Barnard-Brak L, Sulak T, Hatz JK (2011): Macrocephaly in children withautism spectrum disorders. Pediatr Neurol 44:97–100.

61. Davidovitch M, Golan D, Vardi O, Lev D, Lerman-Sagie T (2011): Israelichildren with autism spectrum disorder are not macrocephalic. J ChildNeurol 26:580–585.

62. Fukumoto A, Hashimoto T, Mori K, Tsuda Y, Arisawa K, Kagami S(2011): Head circumference and body growth in autism spectrumdisorders. Brain Dev 33:569–575.

63. Ozgen H, Hellemann GS, Stellato RK, Lahuis B, van Daalen E, Staal WG,et al. (2011): Morphological features in children with autism spectrumdisorders: A matched case-control study. J Autism Dev Disord 41:23–31.

64. Whitehouse AJ, Hickey M, Stanley FJ, Newnham JP, Pennell CE (2011):Brief report: A preliminary study of fetal head circumference growth inautism spectrum disorder. J Autism Dev Disord 41:122–129.

65. Rommelse NN, Peters CT, Oosterling IJ, Visser JC, Bons D, van Steijn DJ,et al. (2011): A pilot study of abnormal growth in autism spectrumdisorders and other childhood psychiatric disorders. J Autism DevDisord 41:44–54.

66. Ververi A, Vargiami E, Papadopoulou V, Tryfonas D, Zafeiriou DI (2012):Clinical and laboratory data in a sample of Greek children with autismspectrum disorders. J Autism Dev Disord 42:1470–1476.

67. Sigurdsson E, Fombonne E, Sayal K, Checkley S (1999): Neurodeve-lopmental antecedents of early-onset bipolar affective disorder. Br JPsychiatry 174:121–127.

68. WHO Expert Committee on Physical Status: The Use and Interpretationof Anthropometry (1995): Physical Status: The Use and Interpretation ofAnthropometry: Report of a WHO Expert Committee. Geneva: WorldHealth Organization.

69. Knickmeyer RC, Gouttard S, Kang C, Evans D, Wilber K, Smith JK, et al.(2008): A structural MRI study of human brain development from birthto 2 years. J Neurosci 28:12176–12182.

70. Baird G, Simonoff E, Pickles A, Chandler S, Loucas T, Meldrum D,Charman T (2006): Prevalence of disorders of the autism spectrum in apopulation cohort of children in South Thames: The Special Needsand Autism Project (SNAP). Lancet 368:210–215.

71. Veena SR, Krishnaveni GV, Wills AK, Kurpad AV, Muthayya S, Hill JC,et al. (2010): Association of birthweight and head circumference atbirth to cognitive performance in 9- to 10-year-old children in SouthIndia: Prospective birth cohort study. Pediatr Res 67:424–429.

72. Golding J, Pembrey M, Jones R (2001): ALSPAC–the Avon LongitudinalStudy of Parents and Children. I. Study methodology. Paediatr PerinatEpidemiol 15:74–87.

73. Cohen MM Jr (2003): Mental deficiency, alterations in performance,and CNS abnormalities in overgrowth syndromes. Am J Med Genet CSemin Med Genet 117C:49–56.

74. Betancur C (2011): Etiological heterogeneity in autism spectrumdisorders: More than 100 genetic and genomic disorders and stillcounting. Brain Res 1380:42–77.

A. Raznahan et al. BIOL PSYCHIATRY 2013;74:563–575 575

75. Stoltenberg C, Schjolberg S, Bresnahan M, Hornig M, Hirtz D, Dahl C,et al. (2010): The Autism Birth Cohort: A paradigm for gene-envi-ronment-timing research. Mol Psychiatry 15:676–680.

76. Kendler KS, Neale MC (2010): Endophenotype: A conceptual analysis.Mol Psychiatry 15:789–797.

77. Hazlett HC, Gu H, McKinstry RC, Shaw DW, Botteron KN, Dager SR,et al. (2012): Brain volume findings in 6-month-old infants at highfamilial risk for autism. Am J Psychiatry 169:601–608.

78. Feingold M, Bossert WH (1974): Normal values for selected physicalparameters: An aid to syndrome delineation. Birth Defects Orig Artic Ser10:1–16.

79. Chiotis D, Tsiftis G, Hatzisymeon M, Maniati-Christidi M, Dacou-Voutetakis A (2003): Height and weight of children of Hellenic originaged 0–18 years (2000–2001): Comparison with data collectedduring the period 1978–1979. Ann Clin Pediatr Univ Atheiensis 50:136–155.

80. Japanese Ministry of Education (2001): Report of the 2000 NationalGrowth Survey of Normal Japanese Children 5-17 Years of Age. Tokyo:Japanese Ministry of Education.

81. Hollingshead AB (1975): Four-Factor Index for Social Status. NewHaven, CT: Yale University Press.

www.sobp.org/journal