systematic review of early intensive behavioral interventions for children with autism

� American Association on Intellectual and Developmental Disabilities 23

VOLUME 114, NUMBER 1: 23–41 � JANUARY 2009 AJIDD

Systematic Review of Early Intensive BehavioralInterventions for Children With Autism

Patricia Howlin and Iliana MagiatiInstitute of Psychiatry, King’s College (London, UK)

Tony CharmanUniversity College, London, Institute of Child Health

AbstractRecent reviews highlight limitations in the evidence base for early interventions for childrenwith autism. We conducted a systematic review of controlled studies of early intensivebehavioral interventions (EIBI) for young children with autism. Eleven studies met inclu-sion criteria (including two randomized controlled trials). At group level, EIBI resulted inimproved outcomes (primarily measured by IQ) compared to comparison groups. At anindividual level, however, there was considerable variability in outcome, with some evi-dence that initial IQ (but not age) was related to progress. This review provides evidencefor the effectiveness of EIBI for some, but not all, preschool children with autism.

DOI: 10.1352/2009.114:23–41

Over the past 2 decades, there has been in-creasing interest in developing effective interven-tions for young children with autism spectrumdisorder. In addition to the many largely untested‘‘alternative’’ therapies (from acupuncture to Zincsupplements; see Research Autism, 2007), a widerange of more soundly based psychoeducationalprograms has been employed. These tend to in-corporate a mix of behavioral, developmental, andeducational approaches (for reviews, see NationalResearch Council, 2001; Scottish IntercollegiateGuidelines Network [SIGN], 2007), and althoughmethods vary, the general goal of such programsis to enhance cognitive, communication, and so-cial skills while minimizing autistic symptomatol-ogy and other problem behaviors. Three mainstrands of intervention have been the focus in themajority of studies conducted to date: programsthat have a specific focus on communication;those in which developmental/educational strate-gies have been employed, and those with a par-ticular emphasis on the use of behavioral princi-ples to improve learning and behavior.

In terms of communication-focused interven-

tions, programs such as the Picture ExchangeCommunication System (Bondy & Frost, 1998)and other alternative/augmentative communica-tion systems were designed in an attempt to pro-vide a communication modality for children whohave no spoken language (see Howlin, 2006).Other programs, designed for both nonverbal andverbal children, use approaches informed by psy-cholinguistic theory to target the early interactionsbetween parents and newly diagnosed children inorder to enhance nonverbal and verbal commu-nication. These include parent communicationtraining approaches (Aldred, Green, & Adams,2004; Drew et al, 2002); More Than Words (Suss-man, 1999), Early Bird (Shields, 2001), and Pre-Linguistic Milieu Therapy (Yoder & Stone, 2006).In addition, there are numerous developmentaland educational programs that combine aspects ofdevelopmental, educational, and behavioral ap-proaches, including Daily Life Therapy (Quill,Gurry, & Larkin, 1989); the Denver Model (Rog-ers et al., 2006); the Douglass Developmental Dis-abilities Center program (Handleman, Harris, Ar-nold, Gordon, & Cohen, 2006); Floor Time/De-

24 � American Association on Intellectual and Developmental Disabilities


Early intensive behavioral interventions P. Howlin, I. Magiati, and T. Charman

velopmental, Individual Difference Relationshipmodel (Greenspan & Wieder, 2003); Son-Rise(Kaufman, 1994); TEACCH (Treatment and Ed-ucation of Autistic and Communication Handi-capped Children; Schopler, 1997), and manymore.

Although all these programs employ some-what different methodologies, they also have im-portant elements in common; for example, tech-niques developed from learning theory (Skinner,1953) are essential components of most. Operantapproaches, as exemplified in the Applied Behav-ior Analysis (ABA) approach (Dunlap, Kern-Dun-lap, Clark, & Robbins, 1991) are particularly fun-damental to behavioral techniques such as PivotalResponse Training (Schreibman & Koegel, 2005),Discrete Trial Training (Maurice, Green, & Luce,1996) and Verbal Behavior (Barbera & Rasmussen,2007) that form part of many modern-day earlyintervention programs for children with autism.

Recently, reviewers of intervention programsfor children with autism have been critical of thestandards of research in this field, pointing to var-ious methodological limitations and failures of ex-perimental design (Levy, Kim, & Olive, 2006;Scottish Intercollegiate Guidelines, 2007; Wheel-er, Blaggett, Fox, & Blevins, 2006). However, thereis also widespread agreement that interventionsbased on ABA, particularly those involving hometherapy and beginning in the preschool years,have been most comprehensively studied and, asa consequence, have the best established evidencebase (Lord et al., 2005; National Research Coun-cil, 2001; Smith, Donahoe, & Davis, 2006).Among the most thoroughly evaluated are pro-grams involving early intensive home-based be-havioral intervention (EIBI). The EIBI approachunderlies the UCLA Young Autism Project, orig-inally developed by Lovaas and his colleagues(1981). Other more recently developed EIBI pro-grams include those described by Maurice, Green,and Luce (1996), Maurice, Green, and Foxx(2001), and Handleman et al. (2006).

Lovaas was among the earliest researchers todemonstrate the effectiveness of behavioral inter-ventions for children with autism and was alsoone of the first to stress the importance of paren-tal involvement in therapy. Proponents of EIBIrecommend that therapy should begin as early aspossible, preferably before the age of 3 years,should take place for approximately 40 hours perweek and last at least 2 years. The programs arehighly prescriptive, with detailed manuals provid-

ed to guide and monitor treatment. Learning ses-sions are provided in a one-to-one discrete trialformat, focusing on the systematic teaching ofmeasurable behavioral units, repetitive practice,and structured presentation of tasks from the mostsimple to the more complex. It has now been 2decades since Lovaas (1987) published his originalpaper on the effectiveness of EIBI, reporting thatpreschool children involved in 1:1 therapy for 40or more hours a week, over at least 2 years,showed major gains (of up to 30 points) in IQ;rates of integration in mainstream school in-creased significantly; and 47% of children weredescribed as attaining ‘‘normal intellectual and ed-ucational functioning (p. 3).’’ A subsequent fol-low-up (McEachin, Smith, & Lovaas, 1993) alsoconfirmed that the gains were maintained intoearly adolescence.

Although in some quarters this study washailed as ‘‘a triumph of behavioral science andbehaviorally scientific clinical application’’ (Baer,1993, p. 373), critics highlighted problems in ex-perimental design (particularly with regard to non-random assignment and intelligence test–retestmeasures) and the practical and financial difficul-ties of replication given the intensity of the pro-gram. It was claims of ‘‘normal functioning,’’however, that gave rise to most contention (Mes-ibov, 1993; Mundy, 1993). This was partly due tothe limited range of assessments employed onwhich to base judgments of ‘‘normality’’ but alsobecause the figure of 47% showing ‘‘normal func-tioning’’ (McEachin et al., 1993, p. 367) repre-sented only 9 out of 19 children. Smith et al.(2001) noted that the 9 children classified as hav-ing normal functioning accounted for most of thegains in IQ. This subgroup showed an average IQincrease of 37 points compared with an averageincrease of 3 points for the other 10 children.Moreover, Shea (2004) pointed out that the pub-lished data from the studies of Lovaas (1987) andMcEachin et al. (1993) indicated that on the in-struments used, only one child had test scores thatwere all within the average range.

The extent of involvement required by par-ents and therapists to follow the UCLA EIBI pro-gram for the recommended number of hours hasrestricted opportunities for replication, and in thefew studies that have been conducted, researchershave generally used a modified version of thismodel. Nevertheless, a number of replications (orat least partial replications) of the original studyhas now been undertaken, and although many do




not report gains of the magnitude reported by Lo-vaas (1987) or McEachin et al. (1993), the resultshave generally been positive.

Although there have been previous reviews ofearly interventions for children with autism (Lordet al., 2005; Rogers, 1998; Smith, Eikeseth, Klev-strand, & Lovaas, 2007), there have been manynew controlled studies in recent years (Rogers’1998 review, for example, appeared before 8 of the11 studies reviewed here were published). Thus,an updated review is required in order to take ac-count of the new evidence reported. In this sys-tematic review, we examined the findings fromcontrolled EIBI studies published in peer-reviewedjournals. Our goal was to examine what conclu-sions can be drawn from research conducted todate and to identify the challenges for future re-searchers.

Method

The following databases were searched forpeer-reviewed papers from 1985 to May 2007:MEDLINE, EMBASE, Cochrane, PSYCHINFO,CINAHL, ERIC, using the terms autism/autistic/autism spectrum disorder and intervention/early inter-vention/behavio(u)ral intervention/EIBI/ABA. Ad-ditional sources of information were the UnitedKingdom National Autistic Society Research Au-tism website; international reviews published bythe New York State Department of Health EarlyIntervention Program (1999); the state of Maine(Maine Administrators of Services for ChildrenWith Disabilities (Maine, 1999); National Re-search Council (2001); National Initiative for Au-tism: Screening and Assessment (2003); Universi-ty of Sydney (Roberts, 2004); New Zealand Min-istry of Health (2006); and National Health Ser-vice (NHS) Scotland (Scottish IntercollegiateGuidelines Network, 2007). The inclusion criteriafor the review were (a) case-control comparisonstudy (not necessarily randomized design) of EIBIfor children with autism; (b) UCLA-affiliated(home- or center-based) or other home-based EIBIprogram largely based on the UCLA model (notnecessarily for 40 hours per week); (c) minimumof 10 participants in EIBI group; (d) age at startof treatment less than 6 years; (e) interventionminimum 12 hours per week; (e) duration mini-mum 12 months; (f) adequate data on IQ or otherstandard measures to allow calculation of extentof gains.

Results

In total, we identified 641 studies that met thesearch term combinations. We excluded studies inwhich the title indicated clearly that the study wasof a single case or a case series without a compar-ison group and studies with a focus on a specificbehavior(s) (e.g., sleeping problems, challengingbehavior), the use of pharmacological interven-tions or other non-ABA programs, or on out-comes for/views of therapists or parents. Abstractsof review papers were checked, and if these werereviews of ABA/EIBI with participants who hadautism their reference lists were further scrutinizedfor possible papers to include. We initially iden-tified 37 papers that appeared to meet criteria forinclusion; 8 of these were overviews of researchin the area or commentaries on previously pub-lished studies; 5 were non-EIBI or a mix of EIBIand other behavioral interventions; 11 had no re-ported IQs and/or no comparison group and/orhad fewer than 10 participants. Thirteen studies(indicated in boldface in the References) met all in-clusion criteria; 2 of them (Eikeseth, Smith, Jahr,& Eldevik, 2007; McEachin et al., 1993) were ex-tensions of previous reports (Eikeseth, Smith,Jahr, & Eldevik, 2002; Lovaas, 1987, respectively).These are presented as single outcome reports (seeTable 1). Only one study (Smith, Groen, &Wynn, 2000) was a fully randomized control trial.Because data were not available for individual par-ticipants, the following summaries/analyses arebased on comparisons of the published groupmeans. All but 4 studies (Eikeseth et al., 2007,2006; Magiati, Charman, & Howlin, 2007; Rem-ington et al., 2007) were carried out in the UnitedStates.

Characteristics of Participants and TherapeuticInterventions

Diagnostic measures varied across studies,with most relying on clinical judgments based onDiagnostic and Statistical Manual of Mental Disor-ders/International Classification of Diseases criteria.Six studies (Cohen, Amerine-Dickens, & Smith,2006; Eikeseth et al., 2002, 2007; Eldevik, Eike-seth, Jahr, & Smith, 2006; Magiati et al., 2007;Remington et al., 2007; Sallows, & Graupner,2005) used the Autism Diagnostic Interview-Revised ADI-R (Lord, Rutter, & le Couteur,1994) to confirm diagnosis. Reported diagnosesincluded children with autism, autism spectrumdisorder, and pervasive developmental disorder.




Tabl

e1.

Sum

mar

yof

Inte

rven

tion

and

Part

icip

ant

Cha

ract

eris

tics

Stud

yD

iagn

osti

cla

bela

nb

Mea

nag

esc

atTi

me

1Ty

peof

stud

yEI

BIco

ndit

ion

Com

pari

son

(C)

cond

itio

n

Lova

as,1

987;

McE

achi

net

al.,

1993

(USA

)D

SM/C

linic

aldi

-ag

nosi

sof

auti

sm

EIBI

�19

C1�

19C2

�21

EIBI

�35

C1�

41C2

�42

Case

cont

rolt

rial

UCL

A-

EIBI

C1:l

ess

inte

nsiv

eth

erap

y,C2

:min

i-m

umth

erap

ySm

ith

etal

.,19

97(U

SA)

DSM

:MR

and

auti

smEI

BI�

11C

�10

EIBI

�36

C�

38Re

tros

pect

ive

case

-co

ntro

lstu

dyU

CLA

-EIB

Im

ulti

site

Min

imum

ther

apy

Shei

nkop

f&

Sieg

el,

1998

(USA

)D

SM:a

utis

m;

PDD

EIBI

�11

C�

11EI

BI�

34C

�35

Retr

ospe

ctiv

eca

se-

cont

rolt

rial

Com

mun

ity-

base

dU

CLA

mod

el

Stan

dard

scho

olin

g

Smit

h,G

roen

,&W

ynn,

2000

(USA

)Cl

inic

diag

nosi

s:au

tism

;PD

DEI

BI�

15C

�13

EIBI

�36

C�

36Ra

ndom

ized

con-

trol

tria

lU

CLA

-EIB

Im

ulti

site

Pare

nt-t

rain

ing/

pub-

licsc

hool

Eike

seth

etal

.,20

02/

2007

(Nor

way

)A

DI-R

auti

smEI

BI�

13C

�12

EIBI

�66

C�

65Ca

se-c

ontr

oltr

ial

Scho

ol-b

ased

UCL

Am

odel

Ecle

ctic

educ

atio

n

Sallo

ws

&G

raup

ner,

2005

(USA

)A

DI-R

auti

smEI

BI�

13C�

10EI

BI�

35C

�37

Rand

omiz

edal

loca

-ti

onto

clin

icor

pare

nt-r

ungr

oup

UCL

A-E

IBI

mul

tisi

tePa

rent

-dir

ecte

dEI

BI

How

ard

etal

.,20

05(U

SA)

DSM

(wit

hse

-ve

rity

deta

ils)

auti

sm;P

DD

EIBI

�29

C1�

16C2

�16

EIBI

�31

C1�

39C2

�35

Case

-con

trol

tria

lCo

mm

unit

y-ba

sed

EIBI

C1:6

auti

smsp

ecifi

ced

.C2:

16ge

neri

csp

ecia

led.

Cohe

net

al.,

2006

(USA

)A

DI-R

auti

sm;

PDD

EIBI

�21

C�

21EI

BI�

30C

�33

d

Case

-con

trol

tria

lU

CLA

-EIB

Im

ulti

site

Ecle

ctic

educ

atio

n

Elde

vik

etal

.,20

06(N

orw

ay)

Mos

tA

DI-R

au-

tism

�LD

EIBI

�13

C�

15EI

BI�

53C

�49

Retr

ospe

ctiv

eca

se-

cont

rolt

rial

Scho

ol/h

ome-

base

dU

CLA

mod

el

Ecle

ctic

educ

atio

n

Mag

iati

etal

.,20

07(U

K)

AD

I-Rau

tism

;A

SDEI

BI�

28C

�16

EIBI

�38

C�

42Ca

se-c

ontr

oltr

ial

Com

mun

ity-

base

dEI

BIA

utis

msp

ecifi

cnu

rs-

ery

Rem

ingt

onet

al.,

2007

(UK

)A

DI-R

auti

smEI

BI�

23C

�21

EIBI

�36

C�

38Ca

se-c

ontr

oltr

ial

Com

mun

ity-

base

dEI

BIM

ixof

ther

apie

s,no

ne1:

1a D

SM�

Dia

gnos

tican

dSt

atist

ical

Man

ual

ofM

enta

lD

isord

ers;

PDD

�pe

rvas

ive

deve

lopm

enta

ldi

sord

er;

AD

I-R

�A

utis

mD

iagn

ostic

Inte

rvie

w-R

evis

ed;

MR

�m

enta

lre

tard

atio

n.bE

IBI

�ea

rly

inte

nsiv

ebe

havi

oral

inte

rven

tions

;C

�co

mpa

riso

ngr

oup

(C1

�G

roup

1;C

2�

Gro

up2

inst

udie

sw

ithm

ore

than

1co

mpa

riso

ngr

oup)

.c In

mon

ths.

dO

nly

age

atdi

agno

sis

give

n.




Smith et al. (1997) and Eldevik et al. (2006) fo-cused specifically on children with intellectual im-pairments; the remainder included children withmean IQs ranging from 50 to 80�.

The mean number of EIBI children per studywas 17.8 (SD � 6.6, range � 11 to 29); the meannumber of comparison cases was 14.7 (SD � 4.1,range � 10 to 21). There was no significant dif-ference in the ages of EIBI and comparison chil-dren at baseline; the mean age of the EIBI chil-dren was 40 months, with a range of 31 to 66.Comparison children’s mean age was 42 months,with a range of 35 to 65. However, it was occa-sionally difficult to establish accurately from thepublished data the age at which intervention ac-tually began, with some researchers (e.g., Cohenet al., 2006) reporting only the ages at which chil-dren were diagnosed or first assessed rather thanthe precise age at which intervention started. Inthe majority of interventions, investigatorsclaimed to follow the behavioral program devel-oped by Lovaas and colleagues as part of theUCLA Young Autism Project; 5 were part of theUCLA multisite replication group; 2 were school-based but affiliated with the UCLA program; 4were community-based programs. Of these,Sheinkopf and Siegel (1998), according to parentalreport, followed the UCLA model; Howard,Sparkman, Cohen, Green, and Stanislaw (2005)and Remington et al. (2007) reported that thera-pists had been trained in a number of ABA/EIBItechniques, including the manualized programs ofMaurice et al. (1996, 2001), discrete trial and in-cidental teaching (Anderson & Romanczyk, 1999),Natural Environment Training (Sundberg & Par-tington, 1999), and Verbal Behavior (Partington &Sundberg, 1998); Magiati et al. (2007) included atleast some consultants trained at UCLA.

The comparison conditions varied from in-tensive, parent-directed (as opposed to clinic-di-rected) intervention (Sallows & Graupner, 2005)to reduced intensity of EIBI program (Lovaas,1987; Smith et al., 1997); eclectic, public school-ing (Cohen et al., 2006; Eikeseth et al., 2002,2006; Eldevik et al., 2006; Sheinkopf & Siegel,1998); specialist autism school (Howard et al.,2005; Magiati et al., 2007) or a mixture of differ-ent interventions (Remington et al., 2007). Twostudies (Howard et al., 2005; Lovaas, 1987) in-volved two separate comparison groups.

Duration and Intensity of InterventionsTable 2 provides information on the approx-

imate duration of intervention and the ages at

which participants were followed-up. Not all pa-pers provided clear information on how long chil-dren were involved in the program, and in somecases children had been enrolled in the therapysome time prior to the initial ‘‘baseline’’ assess-ments. For some studies we were not able to es-tablish the length of time children were actuallyinvolved in intervention. Lovaas (1987) stated thatthe EIBI children received between 2 and 6 yearsor more of intervention. Smith et al., 1997) re-ported that the EIBI children received a minimumof 2 years of therapy; control participants receivedless than 2 years. As can be seen in Table 2, ofthe remaining 9 studies, 3 involved interventionsof less than 24 months; 5 interventions were 24to 36 months, and in 1 (Sallows & Graupner,2005), interventions lasted 4 years. The mean in-tervention time in these 9 studies for both EIBIand comparison groups was (as far as could beestimated) 27.4 months (SD � 10.7, range 14 to48 months).

Actual hours of intervention proved evenmore difficult to estimate. Only a minority of re-searchers (Magiati et al., 2007; Remington et al.,2007; Smith, Groen, & Wynn, 2000) providedtheir own figures, based on parental or therapists’reports, of the total number of hours in therapy.Most gave only a general indication of hours ofintervention (e.g., minimum number of hours perweek); several indicated that time in therapy de-clined after the first year, without providing in-formation on the degree of reduction. In others(e.g., Lovaas, 1987), it was clear that individualvariation was considerable, with some children inthis, the original Lovaas study, receiving over 6years and more than 14,000 hours of therapy,while others in the same study received 40 hoursof therapy per week for 2 years (i.e., approximateestimation of 4,000 hours). Although it was notpossible, with one exception (Smith et al., 2000:M � 2,158 hours, SD � 1,305, range � 1,142 to5,452), to derive an accurate estimate of totalnumber of hours in therapy over the duration oftreatment, most investigators gave an indicationof approximate hours per week, at least during themost intensive phase of intervention. On the basisof the information provided, we estimated thatEIBI children received significantly more hours ofintervention per week than did controls (EIBI M� 29.8 hours per week; SD � 7.6, range 12.5 to40; controls M � 19.1 hours , SD � 8.6, range�10.0 to 31.5), t � 3.03, p � .007.

Despite the limited quality of the available




Tabl

e2.

Sum

mar

yof

Tre

atm

ent

Dur

atio

nan

dIn

tens

ity

Stud

yTr

eatm

ent

dura

tion

Follo

w-u

p(F

U)

cond

ucte

d/M

ean

age

App

roxi

mat

enu

mbe

rof

hour

spe

rw

eek

EIBI

grou

pCo

mpa

riso

ngr

oup

Lova

as,1

987;

McE

achi

net

al.,

1993

From

2to

6�ye

ars

FUaf

ter

2�an

d8�

year

s(a

tap

prox

.age

7ye

ars

&11

.5ye

ars)

40ho

urs

for

2ye

ars,

then

decr

easi

ngLe

ssth

an10

hour

spe

rw

eek

both

com

pari

son

grou

psSm

ith

etal

.,19

97EI

BIa

�2

year

s;Cb

�2

year

sFU

afte

r2.

5to

3ye

ars

Age

:EIB

I71

mon

ths;

C64

mon

ths

(but

som

ech

il-dr

enag

ed9–

10ye

ars)

30ho

urs

per

wee

k�

10ho

urs

per

wee

k

Shei

nkop

f&

Sieg

el,

1998

EIBI

16m

onth

s;C

18m

onth

sFU

:EIB

Iaft

er20

mon

ths;

Caf

ter

18m

onth

sA

ge:E

IBI&

C54

m

27ho

urs

per

wee

k11

hour

spe

rw

eek

Smit

h,G

roen

,&W

ynn,

2000

2–3

year

sFU

afte

rap

prox

.5ye

ars

Age

:EIB

I94

mon

ths;

C92

mon

ths

25ho

urs

per

wee

k1s

tye

ar;

then

redu

cing

15–2

0ho

urs

per

wee

k(p

aren

tan

dsc

hool

base

d)Ei

kese

thet

al.,

2002

,20

07EI

BI31

mon

ths;

C33

mon

ths

FUaf

ter

12m

onth

s&

2.5

to3

year

sA

ge:E

IBI&

C,98

mon

ths

28ho

urs

per

wee

k1s

tye

ar;

18ho

urs

2nd

year

29ho

urs

per

wee

k1s

tye

ar;1

6ho

urs

2nd

year

Sallo

ws

&G

raup

ner,

2005

48m

onth

sFU

afte

r48

mon

ths

Age

:EIB

I&C

83m

onth

s39

hour

spe

rw

eek

1st

year

;37

hour

spe

rw

eek,

2nd

year

32ho

urs

per

wee

k1s

tye

ar;3

1ho

urs

per

wee

k,2n

dye

arH

owar

det

al.,

2005

13–1

4m

onth

sFU

afte

r14

mon

ths

Age

:EIB

I46

mon

ths;

C151

mon

ths;

C249

mon

ths

25–4

0ho

urs

per

wee

kC1

25–3

0ho

urs

per

wee

k;C2

mea

n15

hour

spe

rw

eek

Cohe

net

al.,

2006

36m

onth

s(b

utso

me

EIBI

grou

psha

dup

to40

0ho

urs

EIBI

befo

rein

take

)

FUaf

ter

36m

onth

sA

geat

FUun

clea

r35

–40

hour

spe

rw

eek

15–2

5ho

urs

per

wee

k

Elde

vik

etal

.,20

06EI

BI20

mon

ths;

C21

mon

ths

FUaf

ter

20–2

1m

onth

sA

ge:E

IBI7

3m

onth

s;C

70m

onth

s

12.5

hour

spe

rw

eek

12ho

urs

per

wee

k

Mag

iati

etal

.,20

0724

mon

ths

FUaf

ter

26m

onth

sA

ge:E

IBI6

4m

onth

s;C

68m

onth

s

32ho

urs

per

wee

k26

hour

spe

rw

eek

Rem

ingt

onet

al.,

2007

24m

onth

sFU

afte

r12

&24

mon

ths

Age

:EIB

I&C

4&

5ye

ars

26ho

urs

per

wee

kin

EIBI

�22

/23

child

ren

insc

hool

for

9–13

hour

spe

rw

eek

15–1

7ho

urs

per

wee

kin

scho

ol1s

tye

ar;1

4–22

hour

spe

rw

eek

2nd

year

a Ear

lyin

tens

ive

beha

vior

alin

terv

entio

ns.b

Com

pari

son

grou

p(C

1�

Gro

up1;

C2

�G

roup

2in

stud

ies

with

mor

eth

anon

eco

mpa

riso

ngr

oup)

.




information, we also believed it important to at-tempt an approximate estimate of total hours intherapy in order to assess the potential impact ofintensity on outcome. This estimate was derivedby multiplying Approximate Weeks of Interven-tion � Approximate Hours per Week (allowingfor any reduction in hours reported by authorsafter 1 or 2 years in therapy). The resulting figureswere as follows: EIBI group mean was 3,353 hours(SD � 1,960, range � 1,000 to 7,100); compari-son group mean was 1,980 hours (SD � 1,529,range 836 to 5,952). Although the intensity of in-tervention was considerably less in the compari-son groups than for the EIBI children, this differ-ence fell short of statistical significance, t � 1.78,p � .09, reflecting the very wide range of hoursin both groups.

Baseline and Outcome DataJust as the studies varied with respect to the

characteristics of children involved and the dura-tion and intensity of intervention, there was con-siderable variation in the methods used to assessoutcome. First, not all researchers presented out-come data at the time intervention actuallyceased; they instead provided this information asmeasured at follow-up sometime later. Thus,whereas the mean duration of the interventionprogram itself was 27 months for both EIBI chil-dren and controls (at least in the 9 studies forwhich a calculation of duration was possible), themean time between baseline assessment and fol-low-up was 39.2 months (SD � 30.1, range � 12to �121) for EIBI children and 34.5 months (SD� 19.2, range � 12 to 79) for controls. Althoughthe group difference in time from cessation oftreatment to follow-up was not significant, t �.45, p � .66, even a relatively small time lag fol-lowing the end of intervention further complicat-ed any assessment of the association between du-ration of therapy and degree of improvement.

Differences in the outcome measures used area further factor affecting comparisons amongstudies. Although data on IQ changes over timewere provided in all 11 studies, information onthe instruments used was generally sparse. Thetests used varied, not only from study to study,but from child to child and from baseline to fol-low-up within the same study, and this variabilityin assessments over time can result in spuriousconclusions about the extent of cognitive im-provement (Magiati & Howlin, 2001). In 6 studiesdata for verbal and nonverbal IQ were reported

separately and in 2, investigators reported changesin mental age (MA). Remington et al. (2007) alsocalculated the Reliable Change Index (Jacobson &Truax, 1991) in order to establish, with 95% cer-tainty, that the level of IQ change was not due tomeasurement unreliability/score variability.

The Vineland Adaptive Behavior ScalesVABS (Sparrow, Balla, & Cicchetti, 1984) havebeen used in the more recent studies (Cohen etal., 2006; Eikeseth et al., 2002, 2007; Eldevik etal., 2006; Howard et al., 2005; Magiati et al.,2007; Remington et al., 2007; Sallows & Graup-ner, 2005; Smith et al., 1997). Some authors pre-sented standard scores; others, age equivalents orraw scores (e.g., Remington et al., 2007). The lat-ter, in particular, are difficult to interpret with re-spect to the clinical meaning of any changesfound.

Language data have become better standard-ized over time, with most investigators providinginformation on expressive and language skills asassessed by the Reynell scales, although not allused the same versions of the Reynell (Cohen etal., 2006; Eikeseth et al., 2002, 2007; Eldevik etal., 2006; Howard et al., 2005; Magiati et al.,2007; Remington et al., 2007; Sallows & Graup-ner, 2005; Smith et al., 1997).

Again, although some researchers presentedstandard scores, others cited age equivalents. Ma-giati et al. (2007) reported raw scores because sofew children were able to score above basal and,for the same reason, Remington et al. (2007) sim-ply recorded the number of children able toachieve any score. Few investigators provided ad-equate information on how issues relating to testbasals for standard scores and age equivalents weredealt with. Raw scores are, of course, acceptablewhen children are so far below basal that ageequivalents and standard scores cannot be calcu-lated (Charman, 2004), but when only raw scoresare provided by some authors, and only standardscores or age equivalents by others, comparisonsamong studies become extremely difficult. Includ-ing data on age equivalents, standard scores, andraw scores would be far more informative and aidthe interpretation of any significant group differ-ences. Now that many journals provide access toonline-only appendices, this comprehensive re-porting of data should be possible in the future.

Other outcome data vary to such an extentthat comparisons between studies are not feasible.Some (e.g., Lovaas, 1987, Sallows & Graupner,2005) provided information on personality tests




(e.g., the Personality Inventory for Children: Wirt,Lachar, Klinedinst, & Seat, 1984; Smith et al.,1997); several authors assessed levels of problembehaviors using a range of different measures, in-cluding maladaptive behavior scales, behavioralobservations, or parental reports (Lovaas, 1987;Magiati et al., 2007; Remington et al., 2007;Smith et al., 1997; Smith et al., 2000). In somelater studies (Magiatii et al., 2007; Sallows &Graupner, 2005), researchers also used the ADI-Rto monitor changes in the severity of autismsymptomatology over time (although the AutismDiagnostic Observation Schedule-GenericADOS-G (Lord et al., 2000) might be a better po-tential outcome measure of autism severity (seeAldred et al., 2004; Howlin, Gordon, Pasco,Wade, & Charman, 2007). Most studies providesome limited information on the school status ofparticipants following intervention (e.g., whetherunsupported in mainstream school).

Table 3 presents the information on changesin IQ over time as provided in all studies. The IQdata indicate that in the majority of studies, therewere no significant differences between EIBI andcomparison children at baseline; in one study(Magiati et al., 2007), where there was a groupdifference in IQ (though not MA), this was con-trolled for in the analysis of change. Comparisonof group scores on other tests for which the ma-jority of studies provide information (VinelandComposite; Language Comprehension, and Ex-pression) also indicates no significant group dif-ferences at intake. In the studies reviewed differentstrategies for analysis were used. The most com-mon analysis was a simple t test, ANOVA, orMann Whitney U test on outcome scores (Eike-seth et al., 2002; Lovaas, 1987; Sheinkopf & Sie-gel, 1998; Smith et al., 1997; Smith et al., 2000),including for Smith et al. and Eikeseth et al., one-tailed t tests). Eikeseth et al. (2002, 2007) and Eld-evik et al. analyzed change scores.

In several of the more recent studies, research-ers reported either simple ANCOVAs on outcomescores (Howard et al., 2006; Remington et al.,2007), repeated measures ANOVA (Sallows &Graupner, 2005), or ANCOVA (Cohen et al.,2006; Remington et al., 2007), partly reflectingtheir larger sample size that enabled them to meetmore of the statistical assumptions that such testsrequire. Although these different analyses do notallow a uniform estimate of effect size to be cal-culated from the available group scores, in orderto provide a simple estimate of effect size for IQ,

we calculated EIBI versus comparison group meandifferences divided by the SD of the two groupscombined (Cohen’s d) (see Table 3).

Data on school placement at follow-up wereprovided in the majority of studies (see Table 4).Overall differences, based on study group meansfor IQ, VABS, and language test standard scoresare reported in Table 5 (some data were extrapo-lated from the published MA scores).

At final follow-up, there were significantgroup differences in IQ in 9 of the 11 reports. Ofthe remaining 2 studies, Magiati et al. (2007) re-ported no differences between the EIBI and com-parison groups. Sallows and Graupner (2005) didnot analyze for Group � Time differences; in-stead, they reported IQ improvements in bothgroups while noting considerable variability inboth groups at outcome. Estimated effect sizeswere moderate, �0.60, to large, �0.80, in the ma-jority of studies. However, in other studies show-ing significant results (e.g., Eldevik et al., 2005),the effect size was small, 0.27. The significantgroup difference in this case is due in part to theanalysis of change scores (Time 1 to Time 2) andthe fact that although the group mean score in-creased somewhat for the EIBI group, it decreasedin the comparison group. As is evident from thesummary data shown in Table 5, there was a sub-stantial mean increase in IQs in the EIBI groupsover time. There were also positive changes, of asimilar magnitude, in language scores. The im-provement in VABS scores, although significantin the summary EIBI versus comparison groupanalysis, was relatively small (average 5 standardscore points).

Several additional points should be noted.First, the extent of change reported in the differ-ent studies is considerable, as is clear from Table5. For example, for IQ, investigators in 3 studies(Eldevik et al., 2006; Magiati et al., 2007; Smithet al., 1997) reported changes of less than 10points; 2 (Remington et al., 2007; Sheinkopf &Siegel, 1998), a 10 to 20 points increase; 4 (Cohenet al., 2006; Eikeseth et al., 2002, 2007; Sallows &Graupner, 2005; Sheinkopf & Siegel, 1998), a 21to 30 points increase; and in 2 studies (Howardet al., 2005; Lovaas, 1987), an increase of 31points. The data on language and VABS scoresshow similar variation. Also of note is the varia-tion of the change in mean scores of the compar-ison groups. Although in some studies the meancomparison group IQ increased between baselineand outcome (albeit not as much on average as




Tabl

e3.

Sum

mar

yof

IQs

atB

asel

ine

(Tim

e1)

and

Follo

w-U

p(T

ime

2)by

Gro

ups

Stud

y

EIBI

IQba

se-

line

Tim

e(T

)1

Mea

n(S

D)

CIQ

base

line

T1M

ean

(SD

)EI

BIIQ

follo

w-u

p(F

U)

T2M

ean

(SD

)C

IQFU

T2M

ean

(SD

)

T2(T

3)EI

BIvs

.C

effe

ctsi

zees

tim

ate

Stat

isti

calt

est

Lova

as,1

987;

McE

achi

net

al.,

1993

53.0

[30-

82]a

46.0

[30-

80]

83.3

b(IQ

atsu

bse-

quen

tFU

�84

.5(S

D32

.4))

C152

.2b

(IQat

subs

e-qu

ent

FU�

54.9

,SD

29.1

)C2

57.5

T2no

tca

lcul

ated

d

T3�

0.96

AN

OV

AT1

/T2

&t

test

T1/T

3**

Smit

het

al.,

1997

28.0

(4.9

)27

.0(5

.4)

36.0

(13.

1)24

.0(8

.2)

1.11

tte

st*

Shei

nkop

f&

Sieg

el,

1998

62.8

(27.

4)61

.7(2

0.2)

89.7

(17.

7)64

.3(2

5.0)

1.19

Pair

edt

test

**

Smit

hG

roen

,&W

ynn,

2000

50.5

(11.

2)50

.7(1

3.9)

66.5

(24.

1)49

.7(1

9.7)

0.76

tte

st*

Eike

seth

etal

.,20

02/

2007

61.9

(11.

3)65

.2(1

0.9)

79.1

(18.

1)at

12m

onth

FU;8

6.9

(SD

25.0

)at

fina

lFU

69.5

(18.

4)at

12m

onth

FU;7

1.9

(28.

4)at

fina

lFU

0.53

tte

stT1

/2**

tte

stT1

/3*

Sallo

ws

&G

raup

ner,

2005

50.8

(10.

6)52

.1(8

.9)

73.1

(33.

1)79

.6(2

1.8)

0.23

Not

repo

rted

but

nons

igni

fica

ntbe

twee

ngr

oups

How

ard

etal

.,20

0558

.5(1

8.2)

C153

.7(1

3.5)

C259

.9(1

4.8)

89.9

(20.

9)C1

62.1

(19.

6)C2

68..1

(15.

3)1.

28M

ulti

ple

regr

es-

sion

**Co

hen

etal

.,20

06c

61.6

(16.

4)59

.4(1

4.7)

87[7

2-98

]73

[62-

80]

0.90

eRe

peat

edm

ea-

sure

sA

NCO

VA

f *El

devi

ket

al.,

2006

41.0

(15.

2)47

.2(1

4.7)

49.2

(16.

6)44

.3(1

8.9)

0.27

tte

ston

chan

gesc

ores

***

Mag

iati

etal

.,20

07c

83.0

(27.

9)65

.2(2

6.9)

78.4

(17.

6)65

.3(1

8.0)

0.74

AN

COV

Ac

Rem

ingt

onet

al.,

2007

61.4

(16.

4)62

.3(1

6.6)

68.8

(20.

5)at

1yr

FU;

73.5

(SD

27.3

)at

2yr

FU

58.9

(20.

5)at

1yr

FU;

60.1

(SD

27.8

)at

2yr

FU

0.48

(at

T3)

Repe

ated

mea

-su

res

AN

COV

Af

T1/T

2/T3

**

Not

e.In

thre

est

udie

sIQ

at2n

dfo

llow

-up–

Tim

e3

was

also

repo

rted

.E

IBI

�ea

rly

inte

nsiv

ebe

havi

oral

inte

rven

tions

,C�

com

pari

son,

C1

�G

roup

1;C

2�

Gro

up2

inst

udie

sw

ithm

ore

than

com

pari

son

grou

p.a I

Qra

nge

give

nin

brac

kets

ifSD

not

repo

rted

.bB

lank

indi

cate

sno

SDor

rang

ere

port

ed.

c Diff

eren

cebe

twee

nE

IBI

and

CIQ

cont

rolle

dfo

rby

AN

CO

VA

.dN

oSD

repo

rted

for

T1

orT

2so

not

poss

ible

toca

lcul

ate.

e No

T2

SDre

port

edso

T1

SDs

used

(like

lyto

over

estim

ate

effe

ctsi

zeat

T2)

.f C

ovar

ying

for

Tim

e1

scor

e.*p

�.0

5.**

p�.0

1.**

*p�

.001

.




Table 4. Summary of School Placement at Follow-Up (FU) by Group

StudyEarly intensive behavioral

intervention (EIBI) Comparison

Lovaas, 1987;McEachin et al.,1993

At age 7 & at age 11–12 year FUs, 9/19 children enrolled in regular in-tervention and described as having‘‘normal educational functioning’’

At age 7 FU, 1 child and at age 11–12 no child described as ‘‘normaleducational functioning’’

Smith et al., 1997 All children had initial IQ�40; all re-mained ‘‘very much delayed’’

All children had initial IQ�40; all re-mained ‘‘very much delayed’’

Sheinkopf & Siegel,1998

5 children reported as attendingmainstream, 3 unsupported

No children reported as unsupport-ed in mainstream

Smith, Groen, &Wynn, 2000

4/15 children unsupported in main-stream

No children unsupported in main-stream

Eikeseth et al., 2002,2007

8/13 children continued to receiveEIBI; remainder had full- or part-time support in school, but it isunclear whether this refers to sta-tus at final FU

Details unclear

Sallows & Graupner,2005

8 children unsupported in main-stream but unclear whether theseare in clinic-directed EIBI or com-parison group

Details unclear

Howard et al., 2005 Not reported Not reported

Cohen et al., 2006 6/21 children unsupported in main-stream

1/21 children in mainstream (un/sup-ported)

Eldevik et al., 2006 All children remained intellectuallyimpaired (mean IQ�40)

All children remained intellectuallyimpaired (mean IQ�40)

Magiati et al., 2007 23/28 children continuing with EIBIin conjunction with supportedplacements in mainstream; 5 chil-dren in specialist school provision

All children in specialist educationalprovision

Remington et al.,2007

17/23 children in mainstream formean of 13 hours per week (notknown if un/supported); 5 in spe-cial education; 1 in EIBI

10/21 children in mainstream formean of 22 hours per week (notknown if un/supported); 11 in spe-cial education for mean of 14hours per week

the EIBI group) (e.g., Cohen et al., 2006; Sallows& Graupner, 2005), in the majority of studies, thecomparison group mean IQ remained the same ordecreased slightly.

Second, as noted in almost all reports, al-though group findings were generally positive infavor of the EIBI groups, there was also consid-erable individual change, with a minority of chil-dren showing marked and significant improve-

ments, and some (see Table 4) achieving educa-tional independence (defined as coping without sup-port in mainstream school). For the majority ofchildren, change, although positive, was less dra-matic, and some failed to make progress despite,sometimes, thousands of hours of intensive inter-vention (e.g., Lovaas, 1987). This variability inoutcome is also illustrated by the tendency inmany studies (see Tables 3 and 5) for variance




Tabl

e5.

Sum

mar

yD

ata

From

All

Stud

ies

onC

hang

esin

IQ,V

AB

S,an

dLa

ngua

geSc

ores

Ove

rT

ime

byG

roup

Mea

sure

a

EIBI

b

base

line

Mea

nSD

Com

pari

son

base

line

Mea

nSD

EIBI

follo

w-u

p(F

U)b

Mea

nSD

Com

pari

son

FU

Mea

nSD

EIBI

chan

gesc

ore

Mea

nSD

Rang

e

Com

pari

son

chan

gesc

ore

Mea

nSD

Rang

e

Gro

upch

ange

scor

edi

ffer

ence

Man

nW

hitn

eyZc

IQ(1

1)55

.713

.953

.611

.274

.017

.659

.015

.518

.311

.5�

5.0–

31.5

5.4

9.3

�3.

3–27

.5–2

.27*

VA

BSco

mpo

site

SSd

(8)

61.2

6.4

61.5

6.7

66.2

10.2

56.2

9.3

5.1

6.2

�2.

0–2.

0�

2.4

4.5

�10

.5–5

.8–2

.58*

*Ex

pres

sion

SS(7

)49

.411

.949

.37.

961

.112

.750

.911

.911

.610

.8�

5–5.

01.

78.

5�

7.5–

15.0

–1.7

9Co

mpr

ehen

sion

SS(7

)42

.810

.945

.05.

358

.510

.949

.612

.115

.76.

07.

0–24

.04.

512

.1�

10.8

–27.

0–1

.98*

Not

e.V

AB

S�

Vin

elan

dA

dapt

ive

Beh

avio

rSc

ales

.a N

umbe

rof

stud

ies

prov

idin

gda

tain

pare

nthe

ses.

No

diff

eren

ces

ingr

oup

mea

nsat

base

line.

bD

ata

prov

ided

for

final

FUif

mor

eth

anon

eFU

asse

ssm

ent

repo

rted

.c N

onpa

ram

etri

can

alys

isus

edbe

caus

ech

ange

scor

eshi

ghly

skew

ed.e S

tand

ard

scor

e.*p

�.0

5.**

p�.0

1.

scores in both the EIBI and comparison groupsto increase over time, indicating greater individualvariability over time. As is also evident from Table5, the range of change scores reported was oftensimilar in both EIBI and comparison groups. Im-portantly, these data suggest, in addition to anygroup mean differences, that while some childrenin both groups do well, others do not. However,only a minority of studies (Cohen et al., 2006;Magiati et al., 2007; Remington et al., 2007; Sal-lows & Graupner, 2005; Sheinkopf & Siegel, 1998)present clear data on individual patterns ofchange.

Predictors of OutcomeIndividual variability in response to treatment

resulted in investigators in 7 out of the 11 studiesexploring more systematically variables thatseemed to be related to the extent of change (gen-erally as measured by increases in IQ). Some re-searchers also considered in detail variables thatappeared to distinguish between children with the‘‘best’’ and ‘‘poorest’’ outcomes. Table 6 providesa summary of the intake variables reported as be-ing predictors of favorable outcome, although itis not possible to draw any consistent conclusionsfrom these findings because different variables (orcombinations of variables) were assessed in differ-ent studies. In 4 studies, initial IQ was stronglypositively associated with degree of improvement,but in one (Sallows & Graupner, 2005) initial IQshowed no relationship with outcome. Chrono-logical age was not reported to be associated withoutcome, and although the age range of childrenin these studies was limited (from 3 to 5.5 years),there was nothing in these data to support theclaims of some EIBI proponents that in order tobe effective, intervention must begin by around 3years. Initial language level (particularly receptivelanguage) did appear to be associated with out-come in 4 of the 7 studies in which predictivefactors were explored. Smith et al. (2000) reportedthat children with less severe autism symptom-atology tend to make better progress, but Rem-ington et al. (2007) found that children with high-er levels of behavior problems and autism symp-toms at intake were most likely to show the great-est change.

DiscussionLimitations of the Present Analysis

In order to interpret the true effects of intakevariables on outcome, reliance on the reports of




Table 6. Predictors of Outcome Following Early Intensive Behavioral Interventions (EIBI)

Study Predictors of outcome

Lovaas, 1987; McEachin et al., 1993 Initial IQ/MA associated with positive outcome; initial CA didnot predict

Smith et al., 1997 None reported; all children had initial IQ�40

Sheinkopf & Siegel, 1998 Not reported

Smith, Groen, & Wynn, 2000 Children with milder symptoms (PDDa diagnosis) tended to dobetter than those with diagnosis of autism

Eikeseth et al., 2002/2007 At 1 year follow-up (FU), initial and language level predictive;at 3 year FU, initial IQ weakly predictive

Sallows & Graupner, 2005 Initial imitation, language, VABS Daily Living and Socializa-tion Skills, and ADI-Rb communication best predictors of 3-year outcome. Initial IQ per se did not predict

Howard et al., 2005 Not reported

Cohen et al., 2006c Not reported

Eldevik et al., 2006 Age at intake not a predictor; initial IQ and language com-prehension and expression main predictors

Magiati et al., 2007c Initial IQ and language comprehension best predictors

Remington et al., 2007 Initial IQ, VABS Composite, Communication, and Social do-main scores all positively related outcome; low VABS MotorScores, high behavior problems, higher autism symptomscores at intake, and fewer hours of intervention in Year 2also related to improved outcomes.

aPervasive developmental disorder. bAutism Diagnostic Interview-Revised. cDerived from Figure 1 in published paper.

group findings is no substitute for a detailed meta-analysis, based on standardized scores for individ-ual children from each of the studies assessed. Todate, 11 studies could be identified that met thecriteria for inclusion required for the present re-view and, thus, with cooperation between groups,it is possible that such a meta-analysis could beconducted. Nevertheless, the only variable report-ed consistently across all studies was IQ, and eventhis measure was derived from different tests be-tween and within studies.

Other outcome measures varied across stud-ies, and although better standardization of psy-chometric testing is clearly important, such mea-sures alone do not necessarily provide an adequateglobal picture of a child’s functioning after inter-vention. There is no a-priori reason why changein IQ should be the principal goal of intervention.Thus, a child might show major increases in IQover time without improving his or her ability tofunction in social situations or improving behav-ior or autistic symptomatology. One likely expla-nation of this focus on IQ as the primary out-

come variable in the EIBI literature is that thebehavioral discrete trial training approach is per-haps best suited to teaching concrete skills as com-pared to spontaneous social and communicationbehavior. Few researchers have attempted to assessoutcome in terms of behavioral difficulties/sever-ity of autism or the impact of these variables onfamily life. In contrast, in communication-basedinterventions, investigators have focused on socialand communication outcomes and not IQ (Al-dred et al., 2004; Drew et al., 2002; Howlin et al.,2007; Kasari, Freeman, & Paparella, 2006; Yoder& Stone, 2006). Measures of educational status areunreliable indicators of progress because schoolplacement may depend more on parental prefer-ences and/or local educational practices and pol-icies rather than factors inherent in the child. Sim-ilarly, in attempting to identify factors related totreatment response, most researchers have focusedon easily measurable variables, such as age, initialIQ, or language. The impact of broader variables,such as parental coping ability, family relation-ships, and stress and support networks, has not




been systematically investigated and, indeed, willprove a major methodological challenge.

Treatment fidelity is a further issue requiringgreater consideration. In only a minority of the11 studies did researchers provide data on thequality of the alternative therapies offered, and invery few studies was the comparison interventionautism-specific. Thus, aside from not employingthe EIBI approach, it is likely that the staff mem-bers involved with these children would have lessunderstanding of or expertise in autism and wouldalso be less well-informed about appropriate ed-ucational practice for children with autism in gen-eral. Details of EIBI interventions are also less de-tailed than might be expected, for although reas-surances are given about the high quality controlof these programs, the supporting informationprovided is generally anecdotal. In community-based studies, perhaps, details of treatment fidelityare less important because these are ‘‘effective-ness’’ trials of implementation in real life. How-ever, in many of the studies reviewed, includingthe UCLA-based ones, it was even difficult to pre-cisely establish how many hours of interventionchildren had received. In several investigations, wealso found it difficult to establish whether baselinemeasures had actually been conducted prior to orat the time of the child’s enrollment in the pro-gram. Indeed, there were indications in some stud-ies that children had already been exposed to theintervention for some time previously. Thus, Co-hen et al. (2006) stated that one inclusion criteri-on was that children should have received ‘‘nomore than 400 hours behavioral intervention pri-or to intake’’ (p. 146). Moreover, a few investiga-tors (e.g., Magiati et al., 2007; Sallows & Graup-ner, 2005) mentioned the fact that children in ei-ther or both the EIBI or comparison conditionswere receiving a number of additional, alternativetreatments (e.g., special diets; biological interven-tions such as Secretin injections or megavitamins;or more controversial therapies such as auditoryintegration, cranial osteopathy, or pet therapy) aswell as being involved in extracurricular activities(e.g., music or play therapy). Few, if any, of thesealternative ‘‘treatments’’ have been empiricallytested, with the exception of Secretin, where theevidence is clear that it has no empirical basis asan autism treatment (Williams et al., 2005).

A further issue concerns the optimal durationof EIBI programs. Neither the summary overview(Table 5) nor the data provided in the individualstudies suggests that length of intervention was

related to outcome. Indeed, Remington et al.(2007) noted that children who responded mostpositively to intervention had fewer hours of ther-apy in the second year than those who made theleast progress.

In those studies reporting IQ changes at sev-eral time points, change scores for the period be-tween the first and subsequent follow-up weregenerally much smaller than the changes betweenbaseline and Follow-Up 1 (see Table 7 for a sum-mary of studies with three or more time points).For example, in Lovaas’ (1987) study, the meanincrease in IQ between initial assessment and fol-low-up at 6 to 7 years was 30 points; at the nextfollow-up, when children were 11 to 12 years ofage, the further increase in IQ was 1.5 points. Ei-keseth et al. (2002) found that the first year in-crease was around 17 points, with an 8-point in-crease in the following year. Cohen et al. (2006)reported that mean IQ rose approximately 16points in the first year but then increased byabout 3 to 4 points the following year and about5 points in the final year (data in this study arepresented graphically and, thus, not easy to readprecisely). In Remington et al.’s (2007) study,there was a mean increase in IQ of 8 points inthe first year and a further increase of 4 points inthe subsequent year. Although these subsequentincreases are not negligible, they do suggest thatthe main impact of intervention is in the first year,and, thereafter, increases, at least in IQ, tend toplateau. Of course, IQ is not the only variable ofinterest and, as noted above, is not necessarily themost important indicator of treatment effective-ness. However, in those 3 studies that also showedchanges in VABS scores over several time points,the diminution in progress after the first year waseven more marked, with few changes in standardscores occurring after 12 months (Eikeseth et al.,2002, 2007; Remington et al., 2007; Sallows &Graupner, 2005). Cohen et al. (2006) reported asteady increase in expressive language from 12 to24 months post-intake, but this progressive in-crease was not replicated in the only other study(Remington et al., 2007) in which language scoresover more than two time points were tracked.

Implications for Future Intervention ResearchThe need to improve the quality of autism

intervention research generally has been stressedin a number of recent reviews, with specific rec-ommendations concerning the need to address is-sues such as small sample sizes, nonrandomized




Tabl

e7.

IQC

hang

esin

Early

Inte

nsiv

eB

ehav

iora

lInt

erve

ntio

ns(E

IBI)

Gro

upfo

rSt

udie

sPr

esen

ting

Dat

aO

ver

Thr

eeor

Mor

eT

ime

Poin

ts

Stud

y

Tim

e1

Ave

rage

age

mea

naIQ

(SD

)

Tim

e2

Ave

rage

(app

rox.

)ag

em

ean

IQ(S

D)

Tim

e3

Ave

rage

(app

rox.

)ag

em

ean

IQ(S

D)

Tim

e4

Ave

rage

(app

rox.

)ag

em

ean

IQ

Lova

as,1

987;

McE

achi

net

al.,

1993

3553

b6–

7ye

ars

83.3

b11

.5ye

ars

84.5

(32.

4)N

/Ac

Eike

seth

etal

.,20

02/2

007

6661

.9(1

1.3)

79m

onth

s79

.1(1

8.1)

98m

onth

s86

.9(2

5.0)

N/A

Cohe

net

al.,

2006

30b

61.6

(16.

4)In

itia

lage

�1

year

77c

Init

iala

ge�

2ye

ars

Init

iala

ge�

3ye

ars

87d

Rem

ingt

onet

al.,

2007

3661

.4(1

6.4)

Init

iala

ge�

1ye

ar68

.8(2

0.5)

Init

iala

ge�

2ye

ars

73.5

(27.

3)N

/A

a In

mon

ths.

bSD

not

prov

ided

;ag

eat

diag

nosi

s.c N

otap

plic

able

(onl

yth

ree

time

poin

tsre

port

ed).

dD

eriv

edfr

omFi

gure

1in

publ

ishe

dpa

per.

assignment, inadequate information on the char-acteristics of participants and therapy, and thelimitations of outcome measures (see Charman etal., 2003; Lord et al., 2005; Rogers, 1998). Thevery small number of evaluations conducted byresearchers who had no direct involvement in thetherapy is another source of concern. Althoughthe research base for studies involving behavioralinterventions is, in general, methodologicallystronger than for many other types of interven-tion, it is clear from our review that none of eventhe better controlled EIBI studies met the mostrigorous standards for intervention methodology(National Research Council, 2001). Moreover, dif-ferences in therapeutic intensity, in participants,and in intake and outcome measures have led toconsiderable disagreement and controversyamong researchers and professionals. In turn, thishas resulted in confusion and distress for manyparents of young children with autism spectrumdisorders, who are faced with conflicting claims ofeffectiveness when seeking appropriate treatment.

There is little question now that early inten-sive behavioral intervention is highly effective forsome children. However, gains are not universal,and some children make only modest progresswhile others show little or no change, sometimesafter extremely lengthy periods in treatment. Gur-alnick (2005) noted that one of the most impor-tant tasks in autism early intervention research isto understand why outcomes vary so dramaticallyacross different children, and the crucial questionto be systematically addressed is for which chil-dren is EIBI most and least effective? Unfortu-nately, because of the focus on group differences,existing research provides only limited informa-tion on the outcome for individual children andfew data on moderators or mediators of therapy(Lord et al., 2005; Yoder & Compton, 2004). It isalmost impossible to draw reliable conclusionsabout possible child, family, or environmentalvariables associated with outcome when moststudies involve relatively small numbers of partic-ipants. More detailed, multisite studies with largesamples (see the current United Kingdom Pre-School Autism Communication Trial (2008) for arare exception) and the sharing of data on indi-vidual cases are necessary for such investigations.At the very least, when sample size is small, as iscurrently the case for most treatment evaluationstudies, it is important that more detailed infor-mation on individual differences is presented, inorder to ensure that group differences are not due




mainly to improvements in a small subgroup ofindividuals.

Furthermore, from the papers reviewed here,there is an indication that the immediate impactof EIBI reduces over time, with the first year ofintervention appearing to produce the most sub-stantial gains, at least in IQ. EIBI is a relativelyexpensive form of intervention, and the cost canbe a major deterrent for many parents and foreducation authorities who may be asked to pro-vide funding. The demands on family life are alsoconsiderable. Clearer evidence concerning boththe optimal duration of therapy and the age atwhich it should begin could result in the devel-opment of better targeted, more cost effective pro-grams that could then be made more widely ac-cessible to families. There is also a need to dem-onstrate that EIBI is substantially more effectivethan alternative, high quality autism specific in-terventions, such as specialized preschool provi-sion. In the majority of the studies included inthis review, the alternative intervention has gen-erally been of lower intensity and/or lesser quality(certainly in terms of autism expertise) than theEIBI program to which it is compared. In addi-tion, the failure to control for time in interventionmeans that in some studies the fact that EIBI ap-pears to be more effective than the comparisoncondition could be due simply to differences intreatment intensity, not quality. As noted above,a systematic meta-analysis is required in order toexplore both child and intervention variables thatmay be related to outcome. However, given thevariability of data, it is likely that such an analysisat the present time could only assess the role of avery limited number of variables.

Conclusions and Recommendationsfor Future Research

Since we initially submitted this paper, oneother comprehensive review of early intensive be-havioral interventions based on the UCLA YoungAutism Project model has been published (Rei-chow & Wolery, 2008). Although that analysiswas focused on some different issues and includespapers that did not meet our criteria for inclusion,the principal conclusion parallels our own; that is,there is strong evidence that EIBI is effective forsome, but not all, children with autism spectrumdisorders, and there is wide variability in responseto treatment. The data we present in this reviewalso suggest that if gains are made, they are likely

to be greatest in the first 12 months of interven-tion. It is evident, too, that in recent studies re-searchers have followed some of the recommen-dations arising from earlier reviews (e.g., Lord etal., 2005; Rogers, 1988; Smith et al., 2007), in par-ticular by including a wider range of child mea-sures, especially those related to social and com-munication skills. However, broader measures offamily functioning are rarely included and as yetthere are no data on how these may affect out-come. There remains a dearth of randomized con-trol trials, which are needed in order to provideunbiased evidence of efficacy.

If research and practice in the area of earlyintervention for children with autism are to im-prove, it is clear that increasing the number ofrandomized control trials should be made a highpriority. Nevertheless, because the number of pri-mary outcome measures of such trials is, of ne-cessity, limited, other case control comparisonstudies will continue to be needed but with ad-equate sample sizes in order to explore individualvariables associated with outcome. For such stud-ies to be informative, the following minimum re-quirements need to be adequately addressed: (a)Baseline data on participants should be collectedimmediately prior to, or at the very beginning, ofthe intervention program. (b) The age at whichindividual children begin treatment, the durationand intensity of treatment (measured in hours perweek), and the exact time to follow-up, should bemade explicit. (c) The same data should be avail-able for children in the comparison conditions,with as much additional information as possibleprovided in order to enable readers to assess thetrue quality of that intervention. (d) Data on childvariables need to be better standardized. Assess-ments used should include those currently in gen-eral use (IQ, VABS, language comprehension andexpression), but additional measures of behavioraldisturbance (e.g., the Developmental BehaviourChecklist: Einfeld & Tonge, 1995) are also re-quired. Standard scores, age equivalent scores, andraw scores (at least in an online appendix) shouldbe reported for all assessments. (e) Diagnostic sta-tus should be clearly defined using the ADI-R andADOS-G because these instruments can also beused to monitor changes in autistic symptomatol-ogy over time. (f) Measures of family functioningneed to be included; not only data on marital andeconomic status, but also measures (such as thoseemployed by Remington et al., 2007) of parentalwell-being and support.




Finally, it is important to recognize that au-tism is a highly complex and heterogeneous dis-order characterized by impairments in social be-havior, in communication, and in many aspectsof learning, together with fundamental problemsin acquiring functional, adaptive, and flexible be-haviors. Although behavioral approaches are animportant element of any comprehensive pro-gram (perhaps especially in the early years), otherelements that focus more specifically on social de-velopment and communication will also be re-quired for optimal effectiveness. This will also re-quire more development of appropriate measuresof social and communicative competence to assesschanges in these skills (Lord et al., 2005). More-over, wide variation in child characteristics and inresponsiveness to treatment may require a farmore individualized approach to interventionthan can be delivered using manualized treatmentpackages.

There is good evidence now, from a numberof randomized control trials (Aldred et al., 2004;Howlin et al., 2007; Kasari et al., 2006; Yoder &Stone, 2006), that other, nonintensive interven-tions, particularly those with a focus on commu-nication and joint social interaction, can have asignificant and positive impact on children’s func-tioning. A switch of focus to examining the com-parative effects of interventions with a strong ev-idence base, called equivalence trials, rather thansimply comparing high quality interventions withlow quality/low intensity alternatives, is likely tobe of benefit to many more children and theirfamilies. Assessing what treatments work forwhich children and identifying the individualcharacteristics that predict responsiveness to spe-cific programs and approaches are the challengesthat lie ahead.

References (* indicates reference ishighlighted in tables)Aldred, C., Green, J., & Adams, C. (2004). A new

social communication intervention for chil-dren with autism: Pilot randomized con-trolled treatment study suggesting effective-ness. Journal of Child Psychology and Psychiatry,45, 1420–1430.

Anderson, S. R., & Romanczyk, R. G. (1999). Ear-ly intervention for young children with au-tism: Continuum-based behavioral models.Journal of the Association for Persons with SevereHandicaps, 24, 162–173.

Baer, D. (1993). Quasi random assignment can beas convincing as random assignment. Ameri-can Journal on Mental Retardation, 97, 373–375.

Barbera, M. L., & Rasmussen, T. (2007). The verbalbehavior approach: How to teach children withautism and related disorders. London: JessicaKingsley.

Bondy, A. S., & Frost, L. A. (1998). The PictureExchange Communication System. Seminarsin Speech and Language, 19, 373–389.

Charman, T. (2004). Matching preschool childrenwith autism spectrum disorders and compar-ison children for language ability: Methodo-logical challenges. Journal of Autism and De-velopmental Disorders, 34, 59–64.

Charman, T., & Howlin, P. (with Aldred, C.,Baird, G., degli Espinosa, F., Diggle, T., Kov-shoff, H., Law, J., et al.). (2003). Research intoearly intervention for children with autismand related disorders: Methodological and de-sign issues. Autism, 7, 217–225.

*Cohen, H., Amerine-Dickens, M. S., & Smith,T. (2006). Early Intensive Behavioral Treat-ment: Replication of the UCLA model in acommunity setting. Journal of Developmentaland Behavioral Pediatrics, 27, S2, 145–155.

Drew, A., Baird, G., Baron-Cohen, S., Cox, A.,Slonims, V., Wheelwright, S., Swettenham, J.,Berry, B., & Charman, T. (2002). A pilot ran-domised control trial of a parent training in-tervention for pre-school children withautism Preliminary findings and methodo-logical challenges. European Child and Adoles-cent Psychiatry, 11, 266–272.

Dunlap, G., Kern-Dunlap, L., Clark, S., & Rob-bins, F. (1991). Functional assessment, curric-ular revision, and severe problem behaviors.Journal of Applied Behavior Analysis, 4, 387–397.

Einfeld, S. L., & Tonge, B. J. (1995). The Devel-opmental Behaviour Checklist: The develop-ment and validation of an instrument to as-sess behavioural and emotional disturbance inchildren and adolescents with mental retar-dation. Journal of Autism and DevelopmentalDisorders, 25, 81–104.

*Eikeseth, S., Smith, T., Jahr, E., & Eldevik, S.(2002). Intensive behavioral treatment atschool for 4- to 7-year-old children with au-tism: A one-year comparison controlledstudy. Behavior Modification, 26, 49–68.

*Eikeseth, S., Smith, T., Jahr, E. & Eldevik, S.




(2007). Outcome for children who began in-tensive behavioral treatment between ages 4-7: A comparison controlled study. BehaviorModification, 31, 264–278.

*Eldevik, S., Eikeseth, S., Jahr, E., & Smith, T.(2006). Effects of low intensity behavioraltreatment for children with autism and men-tal retardation. Journal of Autism and Develop-mental Disorders, 36, 211–224.

Greenspan, S. L., & Wieder, S. (2003). Engagingautism: The Floortime Approach to helping chil-dren relate, communicate and think. Jackson,TN: Perseus Books.

Guralnick, M. (2005). Early intervention for chil-dren with intellectual disabilities: Currentknowledge and future prospects. Journal ofApplied Research in Intellectual Disabilities, 18,313–324.

Handleman, J. S., Harris, S. L., Arnold, M., Gor-don, R., & Cohen, M. (2006). The DouglassDevelopmental Disabilities Center: Optionsfor school-age students. In J. S. Handleman,& S. L. Harris (Eds.), School-age education pro-grams for children with autism. The in-betweenyears (pp. 89–113). Austin, TX: Pro-Ed.

*Howard, J. S., Sparkman, C. R., Cohen, H. G.,Green, G., & Stanislaw, H. (2005). A compar-ison of intensive behavior analytic and eclec-tic treatments for young children with autism.Research in Developmental Disabilities, 26, 359–383.

Howlin, P. (2006). Augmentative and alternativecommunication systems for children with au-tism. In T. Charman & W. Stone (Eds.), Socialand communication development in autism spec-trum disorders (pp. 236–266). New York: Guil-ford Press.

Howlin, P., Gordon, K., Pasco, G., Wade, A., &Charman, T. (2007). A group randomised,controlled trial of the Picture Exchange Com-munication System for children with autism.Journal of Child Psychology and Psychiatry, 48,473–481.

Jacobson, N. S., & Truax, P. (1991). Clinical sig-nificance: A statistical approach to definingmeaningful change in psychotherapy re-search. Journal of Consulting and Clinical Psy-chology, 59, 12–19.

Kasari, C., Freeman, S., & Paparella, T. (2006).Joint attention and symbolic play in youngchildren with autism: A randomized con-trolled intervention study. Journal of Child Psy-chology and Psychiatry, 47, 611–620.

Kaufman, B. N. (1994). Son-Rise: The miracle con-tinues. Tiburon, CA: Kramer.

Levy, S., Kim, A.-H., & Olive, M. (2006). Inter-ventions for young children with autism: Asynthesis of the literature. Focus on Autism andOther Developmental Disabilities, 21, 55–62.

Lord, C., Risi, S., Lambrecht, L., Cook, E. H.,Leventhal, B. L., DiLavore, P. C., Pickles, A.,& Rutter, M. (2000). The Autism DiagnosticObservation Schedule-Generic: A standardmeasure of social and communication deficitsassociated with the spectrum of autism. Jour-nal of Autism and Developmental Disorders, 30,205–223.

Lord, C., Rutter, M., & Le Couteur, A. (1994).Autistic Diagnostic Interview–Revised: A re-vised version of a diagnostic interview forcaregivers of individuals with possible perva-sive developmental disorders. Journal of Au-tism and Developmental Disorders, 24, 659–686.

Lord, C., Wagner, A., Rogers, S., Szatmari, P.,Aman, M., Charman, T., et al. (2005). Chal-lenges in evaluating psychosocial interven-tions for autistic spectrum disorders. Journalof Autism and Developmental Disorders, 35,695–711.

*Lovaas, O. I. (1987). Behavioral treatment andnormal educational and intellectual function-ing in young autistic children. Journal of Con-sulting and Clinical Psychology, 55, 3–9.

Lovaas, O. I., Ackerman, A., Alexander, D., Fire-stone, P., Perkins, M., Young, D. B., et al.(1981). Teaching developmentally disabled chil-dren: The ME book. Austin, TX: Pro-Ed.

*Magiati, I., Charman, T., & Howlin, P. (2007). Atwo-year prospective follow-up study of com-munity-based early intensive behavioural in-tervention and specialist nursery provision forchildren with autism spectrum disorders. Jour-nal of Child Psychology and Psychiatry, 48, 803–812.

Magiati, I., & Howlin, P. (2001). Monitoring theprogress of pre-school children with autismenrolled in early intervention programmes:Problems of cognitive assessment. Autism, 5,399–406.

Maine Administrators of Services for ChildrenWith Disabilities (MADSEC). (1999). Reportof the MADSEC Autism Task Force, Man-chester, ME: Author.

Maurice, C., Green, G., & Foxx, R. M. (Eds.),(2001). Making a difference: Behavioral interven-tion for autism. Austin, TX: Pro-Ed.




Maurice, C., Green, G., & Luce, S. (Eds.). (1996).Behavioral intervention for young children withautism: A manual for parents and professionals.Austin, TX: Pro-Ed.

*McEachin, J. J., Smith, T., & Lovaas, O. I.(1993). Long-term outcome for children withautism who received early intensive behavior-al treatment. American Journal on Mental Re-tardation, 97, 359–372

Mesibov, G. B. (1993). Treatment outcome is en-couraging: Comments on McEachin et al.American Journal on Mental Retardation, 97,379–380.

Mundy, P. (1993). Normal versus high-function-ing status in children with autism (Commentson McEachin, Smith, and Lovaas). AmericanJournal on Mental Retardation, 97, 373–391.

National Initiative for Autism: Screening and As-sessment (NIASA). (2003). National AutismPlan for Children (NAPC). Plan for the identifi-cation, assessment, diagnosis and access to earlyinterventions for pre-school and primary schoolaged children with ASD. London: National Au-tistic Society.

National Research Council. (2001). Educating chil-dren with autism (Committee on EducationalInterventions for Children With Autism, Di-vision of Behavioral and Social Sciences andEducation). Washington, DC: National Acad-emy Press.

New York State Department of Health Early Inter-vention Program. (2000). Clinical practice guide-line: Report of the Recommendations for Autism/Pervasive Developmental Disorders: Assessment andintervention for young children (age 0–3 years). Re-trieved October, 10, 2000, from http://www.health.state.ny.us/nysdoh/eip/autism/index.htm

New Zealand Ministry of Health. (2006). Evidencebased guidelines for autism spectrum disorders(Draft report). Wellington, NZ: Author.

Partington, J. W., & Sundberg, M. L. (1998). TheAssessment of Basic Language and LearningSkills: An assessment, curriculum guide andtracking system for children with autism orother developmental disabilities. Danville,CA: Behavior Analysts.

Pre-School Autism Communication Trial (PACT).(2008). Retrieved September 29, 2008, fromhttp://www.medicine.manchester.ac.uk/pact

Quill, K., Gurry, S., & Larkin, A. (1989). Daily lifetherapy: A Japanese model for educating chil-

dren with autism. Journal of Autism and De-velopmental Disorders, 19, 625–635.

Reichow, B., & Wolery, M. (2008). Comprehen-sive synthesis of early intervention behavioralinterventions for young children with autismbased on the UCLA Young Autism Projectmodel. Journal of Autism and DevelopmentalDisorders. DOI 0. 1007/S10803-008-0596-0

*Remington, B., Hastings, R. P., Kovshoff, H., de-gli Espinosa, F., Jahr, E., Brown, T., et al.(2007). Early intensive behavioral interven-tion: Outcomes for children with autism andtheir parents after two years. American Journalon Mental Retardation, 112, 418–438.

Research Autism. (2007). Retriever August 9,2007, from http://www.researchautism.net/pages/welcome/home.ikml

Roberts, J. M. (2004). A review of the research toidentify the most effective models of best practice inthe management of children with autism spectrumdisorders. Centre for Disability Studies, Uni-versity of Sydney, Australia.

Rogers, S. J. (1998). Empirically supported com-prehensive treatments for young childrenwith autism. Journal of Clinical Child Psycholo-gy, 27, 168–179.

Rogers, S. J., Hayden, D., Hepburn, S., Charlifue-Smith, R., Hall, T., & Hayes, A. (2006).Teaching young nonverbal children with au-tism useful speech: A pilot study of the Den-ver model and PROMPT interventions. Jour-nal of Autism and Developmental Disorders, 36,1007–1024.

*Sallows, G. O., & Graupner, T. D. (2005). Inten-sive behavioral treatment for children with au-tism: Four-year outcome and predictors.American Journal on Mental Retardation, 110,417–438.

Schopler, E. (1997). Implementation of TEACCHphilosophy. In D. J. Cohen & F. R. Volkmar(Eds.), Handbook of autism and pervasive devel-opmental disorders. (2nd ed., pp. 767–798).New York: Wiley.

Schreibman, L., & Koegel, R. L. (2005). Trainingfor parents of children with autism: Pivotalresponses, generalization, and individualiza-tion of interventions. In E. D. Hibbs & P. S.Jensen (Eds.), Psychosocial treatment for childand adolescent disorders: Empirically based strat-egies for clinical practice (2nd ed., pp. 605–631).Washington, DC: American PsychologicalAssociation.

Scottish Intercollegiate Guidelines Network




(SIGN). (2007). Assessment, diagnosis, and clin-ical interventions for children and young peoplewith autism spectrum disorders. Edinburgh Scot-land: National Health Service Education forScotland.

Shea, V. (2004). A perspective on the research lit-erature related to early intensive behaviouralintervention (Lovaas) for young children withautism. Autism, 8, 349–367.

*Sheinkopf, S. J., & Siegel, B. (1998). Home-basedbehavioral treatment of young children withautism. Journal of Autism and DevelopmentalDisorders, 28, 15–23.

Shields, J. (2001). The NAS Early Bird pro-gramme: Partnership with parents in early in-tervention. Autism, 5, 49–56.

Skinner, B. F. (1953). Science and human behavior.New York: Macmillan.

Smith, T., Donahoe, P. A., & Davis, B. J. (2006).The UCLA Young Autism Project. In J. S.Handleman & S. L. Harris (Eds.), Preschool ed-ucation programs for children with autism (2nded., pp. 29–48). Austin, TX: Pro-Ed.

Smith, T., Eikeseth, S., Klevstrand, M., & Lovaas,I. (1997). Intensive behavioral treatment forpreschoolers with severe mental retardationand pervasive developmental disorder. Amer-ican Journal on Mental Retardation, 102, 238–249.

*Smith, T., Groen, A. D., & Wynn, J. W. (2000).Randomized trial of intensive early interven-tion for children with pervasive developmen-tal disorder. American Journal on Mental Retar-dation, 105, 269–285

*Smith, T., Scahill, L., Dawson, G., Guthrie, D.,Lord, C., Odom, S., et al. (2007). Designingresearch studies on psychosocial interventionsin autism. Journal of Autism and DevelopmentalDisorders, 37, 354–366.

Sparrow, S. S., Balla, D., & Cicchetti, D. V.(1984). Vineland Adaptive Behavior Scales. Cir-cle Pines, MN: American Guidance Service.

Sundberg, M. L., & Partington, J. W. (1999). Theneed for both discrete trial and natural envi-ronmental language training for children withautism. In P. M. Ghezzi, W. L. Williams, &J. E. Carr (Eds.), Autism behavior analytic per-spectives (pp. 139–156). Reno, NV: ContextPress.

Sussman, F. (1999). More than words: Helping par-ents promote communication and social skills inchildren with autism spectrum disorders. Toronto:Hanen Center.

Wheeler, J. J., Baggett, B. A., Fox, J., & Blevins,L. (2006). Treatment integrity: A review of in-tervention studies conducted with childrenwith autism. Focus on Autism and Other Devel-opmental Disabilities, 21, 45–54.

Williams, K. W., Wray, J. J., & Wheeler, D. M.(2005). Intravenous secretin for autism spec-trum disorder. Cochrane Database SystematicReviews, 20, CD003495.

Wirt, R. D., Lachar, D., Klinedinst, J. K., & Seat,P. D. (1984). Multi-dimensional description ofchild personality: A manual for the PersonalityInventory for Children. Los Angeles: WesternPsychological Services.

Yoder, P. J., & Compton, D. (2004). Identifyingpredictors of treatment response. Mental Re-tardation and Developmental Disabilities ResearchReviews, 10, 162–168.

Yoder, P. J., & Stone, W. L. (2006). Randomizedcomparison of two communication interven-tions for preschoolers with autism spectrumdisorders. Journal of Consulting and ClinicalPsychology, 74, 426–235.

Received 9/3/07, accepted 1/16/08.Editor-in-charge: William E. MacLean, Jr.

Correspondence regarding this article should besent to Patricia Howlin, Department of Psychol-ogy, Institute of Psychiatry, Henry WellcomeBldg., PO 77, De Crespigny Park, London SE58AF, UK. E-mail: [email protected]

systematic review of early intensive behavioral interventions for children with autism

Documents