Article review

J Autism Dev Disord (2012) 42:512–520

DOI 10.1007/s10803-011-1218-9

ORIGINAL PAPER

Overview of Meta-Analyses on Early Intensive Behavioral Intervention for Young Children with Autism Spectrum Disorders

Brian Reichow

Published online: 15 March 2011

� Springer Science+Business Media, LLC 2011

Abstract This paper presents an overview of 5 meta-

analyses of early intensive behavioral intervention (EIBI)

for young children with autism spectrum disorders (ASDs)

published in 2009 and 2010. There were many differences

between meta-analyses, leading to different estimates of

effect and overall conclusions. The weighted mean effect

sizes across meta-analyses for IQ and adaptive behavior

ranged from g = .38–1.19 and g = .30–1.09, respectively.

Four of fve meta-analyses concluded EIBI was an effec-

tive intervention strategy for many children with ASDs. A

discussion highlighting potential confounds and limitations

of the meta-analyses leading to these discrepancies and

conclusions about the effcacy of EIBI as an intervention

for young children with ASDs are provided.

Keywords Early intensive behavioral intervention � EIBI � Early intervention � Autism spectrum disorders � Meta- analysis

Early intensive behavioral intervention (EIBI; sometimes

referred to as intensive behavioral intervention, early

behavioral treatment, Lovaas therapy, etc.) was one of the

frst comprehensive treatment programs for young children

with autism spectrum disorders (ASDs; Lovaas 1981). EIBI

is based on the principles and technologies of applied

behavior analysis and is typically an intensive home-based

program (e.g., intervention lasting 2? years involving

comprehensive programming for upwards of 40 h per week

with an initial emphasis on discrete trial training using

B. Reichow (&) Yale Child Study Center, 230 South Frontage Road, New Haven,

CT 06519, USA

e-mail: brian.reichow@yale.edu

1-to-1 adult-to-child ratios). According to surveys of par-

ents and service providers (Green et al. 2006; Stahmer et al.

2005), EIBI is one of the most common, popular, and

requested treatment approaches for young children with

ASDs.

The frst empirical results of the effects of EIBI were

published in 1987 (Lovaas 1987), and were very encour-

aging; 47% of the children with autism receiving EIBI

achieved best outcome (i.e., post-treatment IQ [ 85 and unassisted placement in a general education classroom or

successful completion of frst grade in a general education

classroom). A follow-up report (McEachin et al. 1993)

suggested much of the gains the children with best outcome

achieved during intervention were maintained for 6 years.

However, the report also revealed some individuals

receiving greater than 7 years of EIBI did not make good

progress. The initial report and subsequent follow-up report

stirred much debate (e.g., Foxx 1993; Gresham and Mac-

Millan 1998; Mesibov 1993; Mundy 1993; Schopler et al.

1989), and many replications ensued (e.g., Birnbrauer and

Leach 1993; Anderson et al. 1987; Cohen et al. 2006;

Sallows and Graupner 2005; Smith et al. 2000). Due much

in part to the strong effects shown in the initial study and

the surrounding debate on the effectiveness of the inter-

vention, EIBI has become the most studied comprehensive

treatment model for young children with ASDs.

Given the large amount of resources invested in EIBI,

precise estimates of the effects of EIBI should be a priority.

Since 2009, fve meta-analyses of EIBI for young children

with ASDs have been published in peer-reviewed journals

(Eldevik et al. 2009; Makrygianni and Reed 2010; Reichow

and Wolery 2009; Spreckley and Boyd 2009; Virués-Ort-

ega 2010). The results and key methodological character-

istics of these fve meta-analyses are shown in Table 1. The

basic fndings of these meta-analyses varied from strong

123

513

T a

b le

1

In cl

u si

o n

c ri

te ri

a an

d m

et a-

an al

y ti

c m

et h

o d

s o

f re

v ie

w s

in cl

u d

ed i

n o

v er

v ie

w

P o

p u

la ti

o n

In

te rv

en ti

o n

s R

es ea

rc h

O

th er

c ri

te ri

a In

cl u

d ed

N

u m

b er

o f

E ff

ec t

W ei

g h

te d

m ea

n

T es

ts o

f E

v id

en ce

o f

(a g

e ra

n g

e)

d es

ig n

s st

u d

ie s

p ar

ti ci

p an

ts

si ze

ef

fe ct

s iz

e (9

5 %

C I)

h

et er

o g

en ei

ty

p u

b li

ca ti

o n

u se

d

b ia

s

E ld

ev ik

e t

al .

A u

ti sm

o r

E IB

I b

as ed

o n

C

o m

p ar

is o

n

1 2

– 3

6 M

o n

th s

o f

9

T re

at m

en t:

1 5

3

S M

D

IQ —

1 .1

0 (

.8 7

– 1

.3 3

) IQ

— Q

(9 )

= 1

0 .0

7 ,

N o

(2 0

0 9

) P

D D

-N O

S

G re

en e

t al

. an

d /o

r co

n tr

o l

tr ea

tm en

t d

u ra

ti o

n

p =

. 3

5

C o

n tr

o l:

1 4

4

A B

— .6

6 (

.4 1

-. 9

0 )

(2 –

7 y

ea rs

) (2

0 0

2 )

g ro

u p

; n

o t

I2 =

1 0

.6 6

ca

se s

tu d

y o

r A

B —

Q (7

) =

8 .5

, se

ri es

p

= .

2 9

I2 =

1 7

.6 5

R ei

ch o

w a

n d

A

u ti

sm ,

A S

D ,

E IB

I b

as ed

o n

R

C T

s, 2

-g ro

u p

[

1 2

m o

n th

s o

f 1

1

T re

at m

en t:

2 5

1

S M

C

IQ —

.6 9

( .3

9 –

1 .0

0 )

IQ —

Q (1

0 )

= 2

2 .6

, Y

es

W o

le ry

P

D D

, P

D D

- L

o v

aa s

U C

L A

co

m p

ar is

o n

, tr

ea tm

en t

d u

ra ti

o n

p

= .

0 2

(2 0

0 9

) N

O S

Y

A P

m o

d el

o

n e

g ro

u p

I2

=

5 1

.2

(\ 7

y ea

rs )

p re

/p o

st

S p

re ck

le y

A

u ti

sm o

r A

p p

li ed

S

y st

em at

ic

R ig

o r—

P E

D ro

4

T

re at

m en

t: 4

1

S M

D

IQ —

.3 8

( -

.0 9

– .8

4 )

IQ —

Q (2

) =

2 .9

9 ,

N o

t

an d

B o

y d

P

D D

b

eh av

io r

re v

ie w

s,

sc o

re C

6

p =

. 2

2

re p

o rt

ed C

o n

tr o

l: 3

5

A B

— .3

0 (

- .1

6 –

.7 7

) (2

0 0

9 )

(1 .5

– 6

y ea

rs )

in te

rv en

ti o

n

R C

T s,

q u

as i-

I2 =

3 3

.1

R C

T s,

A

B —

Q (2

) =

5 .8

7 ,

co n

tr o

ll ed

p

= .

0 5

tr

ia ls

I2

=

3 3

.1

V ir

u és

- A

u ti

sm (

n o

t C

o m

p re

h en

si v

e G

ro u

p d

es ig

n

[ 1

0 h

o f

tr ea

tm en

t p

er

2 2

T

re at

m en

t: 3

2 3

S

M C

a n

d

IQ —

1 .1

9 (

.9 1

– 1

.4 7

) IQ

— I2

= 7

5

Y es

O rt

eg a

sp ec

if ed

) A

B A

w

it h

[ 5

w

ee k

a n

d

S M

D :

C o

n tr

o l:

1 8

0

A B

— 1

.0 9

A

B —

I2 =

6 8

2

0 1

0

in te

rv en

ti o

n

p ar

ti ci

p an

ts

[ 4

5 w

ee k

s o

f to

g et

h er

(.

7 0

– 1

.4 7

) (e

.g .,

M au

ri ce

tr

ea tm

en t

d u

ra ti

o n

et a

l. 2

0 0

1 )

M ak

ry g

ia n

n i

A u

ti sm

, A

S D

, C

o m

p re

h en

si v

e P

re a

n d

p o

st -

H ig

h (

H )

o r

m o

d er

at e

1 4

T

re at

m en

t: 3

0 3

S

M C

a n

d

S M

C :

IQ —

H :

.9 5

S

M C

: IQ

— H

: N

o

an d

R ee

d

P D

D ,

P D

D -

tr ea

tm en

t tr

ea tm

en t

(M )

m et

h o

d o

lo g

ic al

S

M D

: Q

(4 )

= .

5 4

C

o n

tr o

l: 1

6 2

M

: .9

1

(2 0

1 0

) N

O S

b

as ed

o n

A B

A

m ea

su re

s;

ri g

o r

b as

ed o

n

se p

ar at

e M

: Q

(1 0

) =

1 7

.7 3

A B

— H

: .4

2 M

: .4

7

(\ 4

.5 y

ea rs

) ex

cl u

d ed

m

o d

if ed

c ri

te ri

a o

f A

B —

H :

S M

D :

IQ —

H :

.5 7

Q

(3 )

= 7

.9 9

re

v ie

w s

o r

R ei

ch o

w e

t al

. si

n g

le s

u b

je ct

(2

0 0

8 )

M :

.7 3

d

es ig

n s

M :

Q (6

) =

8 .0

3

A B

— H

: .9

7 M

: .6

6

S M

D :

IQ —

H :

Q (2

) =

5 .0

8

M :

Q (7

) =

1 9

.4 3

A B

— H

:

Q (1

) =

4 .3

1

M :

Q (4

) =

1 1

.5 2

123

K ey

: C

I co

n f

d en

ce i

n te

rv al

, P

D D

-N O

S p

er v

as iv

e d

ev el

o p

m en

ta l

d is

o rd

er -n

o t

o th

er w

is e

sp ec

if ed

, E

IB I

ea rl

y i

n te

n si

v e

b eh

av io

ra l

in te

rv en

ti o

n ,

S M

D s

ta n

d ar

d iz

ed m

ea n

d if

fe re

n ce

e ff

ec t

si ze

,

IQ i

n te

ll ig

en ce

q u

o ti

en t,

A B

a d

ap ti

v e

b eh

av io

r, Q

Q -s

ta ti

st ic

, I2

I -s

q u

ar ed

, A

S D

a u

ti sm

s p

ec tr

u m

d is

o rd

er ,

P D

D p

er v

as iv

e d

ev el

o p

m en

ta l

d is

o rd

er ,

U C

L A

Y A

P U

n iv

er si

ty o

f C

al if

o rn

ia a

t L

o s

A n

g el

es Y

o u

n g

A u

ti sm

P ro

g ra

m ,

R C

T s

ra n

d o

m iz

ed c

o n

tr o

l tr

ia ls

, S

M C

s ta

n d

ar d

iz ed

m ea

n c

h an

g e

ef fe

ct s

iz e,

P E

D ro

p h

y si

o th

er ap

y e

v id

en ce

d at

ab as

e sc

al e

o f

q u

al it

y a

ss es

sm en

t, A

B A

a p

p li

ed

b eh

av io

r an

al y

si s,

H h

ig h

m et

h o

d o

lo g

ic al

r ig

o r,

M m

o d

er at

e m

et h

o d

o lo

g ic

al r

ig o

r

J Autism Dev Disord (2012) 42:512–520

514 J Autism Dev Disord (2012) 42:512–520

support of EIBI to a conclusion that EIBI was not superior

to standard care. Four of fve meta-analyses (Eldevik et al.

2009; Makrygianni and Reed 2010; Reichow and Wolery

2009; Virués-Ortega 2010) concluded EIBI was an effec-

tive intervention strategy for many children with ASDs.

For these four meta-analyses, the weighted mean effect

sizes for IQ had a range of g = .57–1.19 and the range of

weighted mean effect sizes for adaptive behavior was

g = .42–1.09, respectively. The one meta-analysis

(Spreckley and Boyd 2009) concluding EIBI was not

superior to standard care reported weighted mean effect

sizes of g = .38 and g = .30 for IQ and adaptive behavior,

respectively.

Although it is evident that much effort was employed by

the research teams conducting the meta-analyses included

in this overview, all meta-analyses had at least one meth-

odological limitation including calculation of effect sizes

based on small samples (sometimes without reference to a

control group), inclusion of non-randomized studies, over

inclusion of participant data, and lack of standardized

comparison or control groups. This paper provides an

overview of the fve meta-analyses on EIBI for young

children with ASDs and an examination of key differences

and potential confounds that might have led to the dis-

crepant fndings.

Overview of Meta-Analyses

Inclusion Criteria of Meta-Analyses

One striking difference across meta-analyses is the varia-

tion in inclusion criteria. Although all meta-analyses syn-

thesized comprehensive treatment programs for young

children with ASDs based on applied behavior analysis, the

specifc defnitions of the intervention varied from restric-

tion to EIBI based on the manuals of Lovaas (e.g., Lovaas

1981, 2003; inclusion criterion of Reichow and Wolery

2009) to broader defnitions of EIBI leading to the inclu-

sion of programs such as Pivotal Response Treatment and

Group Intensive Family Training (studies by Baker-Ericzen

et al. 2007 and Anan et al. 2008, respectively; inclusion

criterion of Virués-Ortega 2010), which have signifcant

differences from the treatment described in the Lovaas

manuals and many conceptualizations of EIBI. Care must

be taken when conducting meta-analyses not to combine

studies evaluating different independent variables (fre-

quently referred to in meta-analysis as the apples and

oranges problem; Borenstein et al. 2009), which appears to

be a possible confound that cannot be ruled out of all meta-

analyses included in this overview. An additional limitation

of each meta-analysis is that all of the studies on EIBI had

at least one methodological shortcoming including use of

quasi-experimental designs, small sample sizes, non-ran-

dom assignment to groups, inadequate participant charac-

terization, narrow and inadequate outcome measures, lack

of fdelity data, and lack of standardized treatment methods

for control and/or comparison groups.

The different defnitions of EIBI was largely responsible

for the differences in which studies were included in each

meta-analysis, which resulted in large differences in the

total number of studies within each meta-analysis (from 4

studies with a total of 41participants in the Spreckley and

Boyd (2009) meta-analysis to 22 studies with a total of 323

participants in the Virués-Ortega (2010) meta-analysis).

Table 2 provides characteristics of each study included

across meta-analyses. As shown in Table 2, most studies

were included in two, three, or four meta-analyses, except

for one study that was included in all fve meta-analyses

(Smith et al. 2000) and eight studies that were included in

only one meta-analysis apiece (Anan et al. 2008; Baker-

Ericzen et al. 2007; Ben-Itzchak et al. 2008; Harris et al.

1991; Harris and Handleman 2000; Boyd and Corley 2001,

Matos and Mustaca 2005, and Reed et al. 2007b).

The other inclusion criterion likely to have had a sig-

nifcant effect on the conclusions of each meta-analysis is

the research design. In the evaluation of EIBI, studies

using multiple research designs (e.g., randomized clinical

trials, retrospective pre/post comparisons, multiple-arm

trials) with many different types of comparison groups

(e.g., standardized nursery school, treatment as usual,

eclectic models) have been conducted. No meta-analysis

restricted inclusion to randomized control trials, which is

a common recommendation in meta-analysis (Reeves

et al. 2008). All meta-analyses restricted inclusion to

group research design studies. Eldevik et al. (2009) and

Spreckley and Boyd (2009) further restricted the criteria

to comparative group research designs, which limited the

number of studies meeting eligibility criteria. Given the

lack of a standardized comparison group, both strategies

seem justifable, however, the inclusion of non-random-

ized studies is a potential confound and should be con-

sidered a limitation of all meta-analyses included in this

overview. The lack of standardized conditions for com-

parison groups creates a situation in which drawing strong

conclusions about the effectiveness of EIBI is diffcult and

should be considered a limitation that needs to be care-

fully addressed in future meta-analyses. Furthermore, the

lack of a standardized comparison group across studies

evaluating EIBI also created a situation in which the

research teams conducting meta-analysis had to make

decisions on how to interpret different comparison groups,

which had signifcant consequences on the outcome of

each meta-analysis.

123

515 J Autism Dev Disord (2012) 42:512–520

Differences in the Interpretation of Comparison Groups

Misinterpretation of Sallows and Graupner Parent-

Directed EIBI Group

The interpretation of a comparison group had a signifcant

impact on the outcome of the Spreckley and Boyd (2009)

meta-analysis, which interpreted the parent-directed EIBI

group of the Sallows and Graupner (2005) study as a

control group. In the Sallows and Graupner study, partic-

ipants in the parent-directed EIBI group received greater

than 30 h of EIBI per week using the same curriculum

(Lovaas 1981; Maurice et al. 1996) delivered from thera-

pists hired from the same agency as the clinic-directed

EIBI group, which also received greater than 30 h of EIBI

per week. The treatment received by the participants in the

parent-directed EIBI group was not equivalent to standard

care or a traditional no-treatment control group and should

not be considered as such (see Smith et al. 2009 for further

explanation). In fact, the results showed that, on average,

both the parent-directed EIBI group and the clinic-directed

EIBI group made signifcant gains on the standardized

assessments between pre-treatment and post-treatment but

there were not statistically signifcant differences between

the two EIBI groups. The interpretation of the Sallows and

Graupner parent-directed EIBI group as a control group

likely led to the smaller effect sizes found in the Spreckley

and Boyd meta-analysis and their subsequent conclusion

that EIBI was not superior to standard care.

It is noteworthy that the Spreckley and Boyd (2009)

meta-analysis was the only meta-analysis included in this

overview to calculate an effect size for the Sallows and

Graupner (2005) study with the parent-directed EIBI group

treated as a control group. Eldevik et al. (2009) excluded

the Sallows and Graupner study because they concluded

the study did not have a control or comparison group. The

other three meta-analyses included in this overview

(Makrygianni and Reed 2010; Reichow and Wolery 2009;

Virués-Ortega 2010) calculated the standardized mean

change effect size for the Sallows and Graupner study,

which is calculated with respect to change scores and not

post-treatment differences between groups. Although the

Sallows and Graupner parent-directed EIBI group was

treated as a control group in only one of fve meta-analyses

included in this overview, it has occurred elsewhere with

similar consequences. Multiple health insurance agencies

(e.g., Aetna 2010; Blue Cross and Blue Shield 2009; Cigna

2009) have either made a similar misinterpretation or used

the Spreckely and Boyd results to conclude the effective-

ness of EIBI has not been well established, leading to

policy decisions denying coverage of the treatment. The

misinterpretation of the parent-directed EIBI group of the

Sallows and Graupner study as a control group (and lim-

itation to randomized clinical trials) also likely led to

erroneous conclusions in the recent What Works Clear-

inghouse Intervention Report: Lovaas Method of Applied

Behavior Analysis (What Works Clearinghouse 2010). The

signifcance of the misinterpretation of the Sallows and

Graupner parent-directed EIBI group cannot be under-

stated and future reviews should take great care to ensure

this mistake is not made.

Multiple-arm Studies

Multiple-arm studies compare one group of participants

receiving a treatment (e.g., EIBI) to at least two other

groups not receiving that treatment (e.g., TAU and no

treatment control). Three studies included in one or more

meta-analyses were conducted using multiple-arm meth-

odology (Howard et al. 2005; Lovaas 1987; Reed et al.

2007b). When a multi-arm trial is included in a meta-

analysis, recommended practice suggests using only one

comparison either by selecting the comparison that is the

closest to other comparisons in the meta-analysis or by

creating a comparison that averages the results of all pair-

wise comparisons between the treatment and comparison

groups (Borenstein et al. 2009; Higgins et al. 2008). It

appears that most meta-analyses including multiple-arm

trials (Makrygianni and Reed 2010; Reichow and Wolery

2009; Virués-Ortega 2010) followed these conventions.

However, the Eldevik et al. (2009) meta-analysis has

multiple effect size estimates for the treatment group of the

Howard et al. (2005) study, creating a situation in which

the results of the participants of the treatment group

counted twice. Given the large effect size estimates of both

comparisons from the Howard et al. study, it is possible

that the inclusion of multiple comparisons infated the

weighted mean effect sizes and should therefore be con-

sidered a limitation.

Effect Size Calculations

Because studies with and without comparison groups were

included across meta-analyses, two different types of effect

size estimates were used. The standardized mean difference

effect size with Hedges and Olkin’s (1985) small sample

correction, which compares post-treatment scores for the

treatment and comparison groups, could be calculated for

studies comparing one group receiving EIBI with another

group not receiving EIBI. For studies without a comparison

group, the standardized mean change effect size with

Hedges and Olkin’s small sample correction, which com-

pares pre-treatment and post-treatment scores of one group,

had to be used. Across meta-analyses calculations based on

123

516 J Autism Dev Disord (2012) 42:512–520

Table 2 Characteristics of studies included in reviews

Study Year Included in Pretreatment participant characteristics by group Treatment characteristics

Group n Age M,F IQ VABS EL RL Model hr/wk Mo of Tx

Lovaas 1987 E, R, V, M TX 19 34.6 – 62.7 – – – UCLA 40 24?

C 19 40.9 – 57.0 – – – UCLA 10 24?

C 21 \42 – 60.0 – – – TAU – 24?

Anderson et al. 1987 R, V, M TX 14 42.8 – 57.3 50.7 UCLA 15–25 12–24

Harris et al. 1991 V TX 9 50.1 8,1 67.6 – – – EIBI 35–45 11.4

Birnbrauer and Leach 1993 E, R, V TX 9 38.1 5,4 51.3 46.1 – – UCLA 18.7 21.6

C 5 33.2 5,0 54.5 51.5 – – – – 24

Smith et al. 1997 E, R, V, M TX 11 36 11,0 28 50.3 – – UCLA 30 35

C 10 38 8,2 27 – – – UCLA 10 26

Sheinkopf and Siegel 1998 R, V TX 11 33.8 – 62.8 – – – UCLA 27.0 15.7

C 11 35.3 – 61.7 – – – TAU 11.1 18

Weiss 1999 V, M TX 20 41.5 19,1 – 49.9 – – EIBI 40 24

Harris and Handleman 2000 V TX 27 49.0 – 59.3 – – – EIBI 35–40 93

Smith et al. 2000 E, R, S, V, M TX 15 36.1 12,3 50.5 63.4 41.9 37.3 UCLA 24.5 33.4

C 13 35.8 11,2 50.7 65.2 45.6 38.3 UCLA 15–20 24

Bibby et al. 2002 R, V TX 66 45.0 55,11 50.8 54.5 – – UCLA 30.3 32.8

Boyd and Corley 2001 R TX 22 41.3 16,6 – – – – UCLA 20–30 23

Eikeseth et al. 2002 E, S, V TX 13 66.3 8,5 61.9 55.8 45.1 49.0 UCLA 28.0 12.2

C 12 65.0 11,1 65.2 60.0 51.2 50.4 Eclectic 29.1 13.6

Howard et al. 2005 E, V, M TX 29 30.9 25,4 58.5 70.5 51.9 52.2 EIBI 25–40 14.2

C 16 37.4 13,3 53.7 69.8 43.9 45.4 Eclectic 25–30 13.3

C 16 34.6 16,0 59.9 71.6 48.8 49.0 Eclectic 15 14.8

Matos and Mustaca 2005 V TX 9 48 8,1 31 21 – 32 UCLA 30 9–12

Sallows and Graupner 2005 R, S, V, M TX 13 35.0 11,2 50.9 59.5 47.9 38.9 UCLA 37.6 48

TX 10 37.1 8,2 52.1 60.9 48.4 38.8 UCLA 31.3 48

Cohen et al. 2006 E, R, V, M TX 21 30.2 18,3 61.6 69.8 52.9 51.7 UCLA 35–40 36

C 21 33.2 17,4 59.4 70.6 52.8 52.7 Eclectic – –

Eldevik et al. 2006 E, R, V, M TX 13 53.0 10,3 41.0 52.5 33.8 37.3 UCLA 12.5 20.3

C 15 49.0 14,1 47.2 52.5 41.6 33.2 Eclectic 12.0 21.4

Baker-Ericzen et al. 2007 V TX 158 49.4 128,28 – – – – PRT – 12

Ben-Itzchak and Zachor 2007 V, M TX 25 26.6 23,2 70.7 – – – EIBI 35 12

Eikeseth et al. 2007 R, S TX 13 66.3 8,5 61.9 55.8 45.1 49.0 UCLA 28.0 31.4

C 12 65.0 11,1 65.2 60.0 51.2 50.4 Eclectic 29.1 33.3

Magiati et al. 2007 R, V, M TX 28 38.0 27,1 83.0 59.6 2.2r 4.9r UCLA 32.4 25.5

C 16 42.5 12,4 65.2 55.4 1.7 r 2.9 r Eclectic 25.6 26.0

Reed et al. 2007a V, M TX 12 40 11,1 56.8 58.2 – – EIBI 30.4 9

C 20 43 18,2 57.8 53.0 – – Eclectic 12.7 9

C 16 38 – 53.4 58.6 – – Portage 8.5 9

Reed et al. 2007b M TX 14 42.9 14,0 60.1 59.3 – – EIBI 30.4 9–10

C 13 40.8 13,0 56.6 56.5 – – EIBI 12.6 9–10

Remington et al. 2007 E, V, M TX 23 38.4 – 61.4 114.8r – – EIBI 25.6 24

C 21 35.7 – 62.3 113.6r – – TAU 15.3 24

Anan et al. 2008 V TX 72 44 61,11 51.7 53.11 – – EIBI 15 2.8

Ben-Itzchak et al. 2008 V TX 44 27.3 43,1 74.8 – – – EIBI 45 12

C 37 24.2 23,14 71.0 – – – TAU – 12

Key: Age average age by group in months, M male, F female, IQ intelligence quotient, VABS vineland adaptive behavior scales (Sparrow et al. 1984) composite standardized score, EL expressive language, RL receptive language, hr/wk average number of hours per week of treatment, mo of tx average number of months of treatment, E Eldevik et al., 2009, R Reichow and Wolery 2009, V Virués-Ortega 2010, M Makrygianni and Reed 2010, S Spreckley and Boyd 2009, TX treatment group, C control/comparison group,—not reported UCLA University of California at Los Angeles, TAU treatment as usual, EIBI early intensive behavioral intervention, PRT pivotal response treatment, r raw score

123

517 J Autism Dev Disord (2012) 42:512–520

the different effect sizes led to very large differences (e.g.,

[g = 1.50) in effect size estimates for individual studies A sensitivity analysis from one meta-analysis (Virués-Ortega

2010) including both effect sizes suggested that studies

with control groups had a larger weighted mean effect size

for IQ but a smaller weighted mean effect size for adaptive

behavior than studies that did not contain control groups.

Because the standardized mean change effect size does not

account for maturation, the use of this effect size in meta-

analysis should be considered a potential confound and

limitation of the meta-analyses using this estimate.

Moderator Analyses

Three reviews (Makrygianni and Reed 2010; Reichow and

Wolery 2009; Virués-Ortega 2010) concluded their meta-

analysis showed enough between group differences to con-

duct moderator analyses. Makrygianni and Reed used partial

correlations controlling for methodological quality to

examine seven treatment and pre-intervention child char-

acteristics (treatment intensity, treatment duration, parental

training, chronological age, IQ, language, and adaptive

behavior). They found large relations suggesting (a) higher

treatment intensity was related to larger changes and greater

between group differences in IQ and adaptive behavior,

(b) greater treatment duration was related to greater between

group differences in adaptive behavior, (c) inclusion of

parent training was related to greater between group differ-

ences in adaptive behavior, and (d) better pre-treatment

adaptive behavior was related to larger changes in language

and greater between group differences in adaptive behavior.

No statistically signifcant relations were found for pre-

intervention chronological age, IQ, or language ability.

Reichow and Wolery (2009) used analysis of variance

methods to examine methodological rigor and method of

group assignment, both of which did not have a statistically

signifcant relation to changes in IQ. Weighted multiple

regression was used to examine six additional variables

(model of supervisor training, treatment density [intensity],

treatment duration, total hours of treatment, pre-intervention

chronological age, and pre-intervention IQ). Only one vari-

able had a signifcant relation; studies in which supervisors

were trained using the UCLA procedures had greater

increases in IQ scores.

Finally, Virués-Ortega (2010) used random-effects

meta-regression models and dose–response meta-analysis

to examine relations between effect sizes for IQ, language

composite, and adaptive behavior and intervention (dura-

tion and intensity) and child (pre-intervention age and pre-

intervention IQ) characteristics. The meta-regression

showed longer treatment duration was related to larger

differences in language composite scores. The dose–

response meta-analysis suggested longer treatment duration

was related to higher expressive and receptive language

scores and greater treatment intensity was related to higher

adaptive behavior scores.

Publication and Selection Bias

Publication bias should be considered a potential confound

in all of the meta-analyses. All meta-analyses included in

this overview only included studies published in peer-

reviewed journals, which increases the threat of publication

bias. Publication bias was assessed in four meta-analyses;

two analyses found evidence of publication bias (Reichow

and Wolery 2009; Virués-Ortega 2010) and two did not

(Eldevik et al. 2009; Makrygianni and Reed 2010).

Therefore, it is unclear what, if any, effect publication bias

might have had on the results of these meta-analyses;

future meta-analyses should consider more inclusive

inclusion criteria. A related risk is selection bias. Three of

fve meta-analyses (i.e., Eldevik et al. 2009; Makrygianni

and Reed 2010; Spreckley and Boyd 2009) included in this

review were conducted by a research team that included at

least one individual previously involved in studying EIBI.

Furthermore, the author of this overview was involved in

one of the meta-analyses included in this review (Reichow

and Wolery 2009). Although peer-review might help limit

the threat of selection bias, it cannot be ruled out.

Conclusions and Future Recommendations

This paper presents an overview of fve meta-analyses on

EIBI for young children with ASDs. By synthesizing the

results across multiple studies, meta-analysis can be a

powerful tool for estimating the average effects of an

intervention; thus, the collective and accumulating evi-

dence supporting EIBI from meta-analytic studies cannot

be dismissed. On average, EIBI can be a powerful inter-

vention capable of producing large gains in IQ and/or

adaptive behavior for many young children with ASDs.

Despite their differences, most (4 of 5) meta-analyses

(Eldevik et al. 2009; Makrygianni and Reed 2010; Reichow

and Wolery 2009; Virués-Ortega 2010) reached the con-

clusion that EIBI is an effective intervention. It should be

noted that the four meta-analyses reaching this conclusion

are also the four meta-analyses that properly interpreted the

Sallows and Graupner (2005) parent-directed EIBI group.

Stated differently, all meta-analyses correctly interpreting

the Sallows and Graupner parent-directed EIBI group

concluded EIBI is an effective intervention. The conclusion

that EIBI can be an effective intervention for many chil-

dren with autism is also supported by multiple descriptive

reviews (e.g., Granpeesheh et al. 2009; Eikeseth 2009;

Matson and Smith 2008; Rogers and Vismara 2008) and in

123

518 J Autism Dev Disord (2012) 42:512–520

a recent ‘‘mega-analysis’’ of 309 individual participant data

(Eldevik et al. 2010). Furthermore, the current evidence on

the effectiveness of EIBI meets the threshold and criteria

for the highest levels of evidence-based treatments across

defnitions (e.g., Kratochwill and Stoiber 2002; National

Autism Center 2009; Odom et al. 2005; Reichow 2011;

Silverman and Hinshaw 2008). Collectively, EIBI is the

comprehensive treatment model for individuals with ASDs

with the greatest amount of empirical support and should

be given strong consideration when deciding treatment

options for young children with ASDs.

Although the average effects of EIBI appear to be strong

and robust, no treatment, including EIBI, has been effective

for all children with ASDs. Therefore, data providing

information on the child characteristics that are most likely

to be associated with best outcomes are needed. Because of

the discrepant fndings across moderator analyses, the

meta-analyses included in this review shed little light on

this issue. To continue to move the feld forward and

increase the knowledge on effective treatments for children

with ASDs, it is imperative that thorough pre-treatment

participant characterization and collection of outcome data

across a broad range of measures be collected.

Discrepancies across moderator analyses were also seen

with respect to treatment characteristics, suggesting the

specifc treatment components with the greatest effects

remain unclear. Recent survey data suggest that while EIBI

programs often use a similar conceptual foundation (e.g.,

intensive intervention based on applied behavior analysis).

specifc program characteristics vary across and within

programs (Love et al. 2009). To fully realize the potential

benefts of EIBI, additional knowledge on the characteris-

tics of EIBI programs outside of treatment studies (i.e.,

how EIBI is used in real world settings) is needed.

Guidelines focusing on the intensity, duration, level of

treatment fdelity, and therapist experience and/or training

necessary to achieve optimal outcomes should also be more

closely measured and reported in future research.

Finally, in addition to the greater specifcity needed for

treatment components, better knowledge about treatment

outcomes are needed. Based on the meta-analyses reviewed

for this overview, most young children with ASDs

receiving EIBI can expect, on average, large increases in

IQ and lesser (but still signifcant) increases in adaptive

behavior. Although some studies have shown large gains

on standardized measures of language (e.g., Cohen et al.

2006; Smith et al. 2000), it has been less frequently

reported in studies of EIBI and was not synthesized in all of

the meta-analyses. A better understanding of the effects of

EIBI on language abilities is needed. Moreover, although

measuring social competence is diffcult, it is a defning

feature of ASDs (Kanner 1943) and determining the effects

of this intervention, if any, on this core feature of the

disorder should be given careful consideration in future

studies. Finally, the differences in measurement instru-

ments across studies for psychopathology have likely led to

no synthesis of these measures. Better measurement and

reporting of psychopathology and in turn, a better under-

standing of the effects of EIBI on the core symptomotology

of children receiving the treatment should be a priority.

Data refecting typical changes on standard outcomes such

as IQ, adaptive behavior, language abilities, and psycho-

pathology due to treatment in real life settings should also

be collected. Once data on optimal child characteristics,

necessary treatment components, and likely outcomes are

collected, parents and clinicians will be able to make more

informed choices when selecting EIBI as a treatment for

young children with ASDs.

References

Studies included in the reviews are indicated with an

asterisk (*). Reviews included in the overview are

indicated with a cross (1)

Aetna. (2010). Clinical policy bulletin: Pervasive developmental

disorders. Obtained July 1, 2010 from http://www.aetna.com/

cpb/medical/data/600_699/0648.html.

*Anan, R. M., Warner, L. J., McGillivary, J. E., Chong, I. M., &

Hines, S. J. (2008). Group intensive family training (GIFT) for

preschoolers with autism spectrum disorders. Behavioral Inter- ventions, 23, 165–180.

*Anderson, S. R., Avery, D. L., DiPietro, E. K., Edwards, G. L., &

Christian, W. P. (1987). Intensive home-based early intervention

with autistic children. Education and Treatment of Children, 10, 353–366.

*Baker-Ericzen, M. J., Stahmer, A. C., & Burns, A. (2007). Child

demographics associated with outcomes in a community-based

pivotal response training program. Journal of Positive Behav- ioral Interventions, 9, 52–60.

*Ben-Itzchak, I. E., Lahat, E., Burgin, R., & Zachor, A. D. (2008).

Cognitive, behavior and intervention outcome in young children

with autism. Research in Developmental Disabilities, 29, 447–458.

*Ben-Itzchak, E., & Zachor, D. A. (2007). The effects of intellectual

functioning and autism severity on outcome of early behavioral

intervention for children with autism. Research in Developmen- tal Disabilities, 28, 287–303.

Blue Cross and Blue Shield. (2009). Special report: Early intensive

behavioral intervention based on applied behavior analysis

among children with autism spectrum disorders. Obtained July

1, 2009 from: http://www.bcbs.com/blueresources/tec/vols/23/

23_09.pdf.

*Bibby, P., Eikeseth, S., Martin, N. T., Mudford, O. C., & Reeves, D.

(2002). Progress and outcomes for children with autism receiv-

ing parent-managed interventions. Research in Developmental Disabilities, 23, 81–104.

*Birnbrauer, J. S., & Leach, D. J. (1993). The Murdoch early

intervention program after 2 years. Behaviour Change, 10(2), 63–74.

123

519 J Autism Dev Disord (2012) 42:512–520

Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H.

(2009). Introduction to meta-analysis. Hoboken, NJ: Wiley. *Boyd, R. D., & Corley, M. J. (2001). Outcome survey of early

intensive behavioral intervention for young children with autism

in a community setting. Autism, 5, 430–441. Cigna. (2009). Cigna medical coverage policy: Intensive behavioral

interventions. Obtained July 1, 2010 from: http://www.cigna.ca/

customer_care/healthcare_professional/coverage_positions/medical/

mm_0499_coveragepositioncriteria_intensive_Behavioral_interventions.

pdf.

*Cohen, H., Amerine-Dickens, M., & Smith, T. (2006). Early

intensive behavioral treatment: Replication of the UCLA model

in a community setting. Developmental and Behavioral Pediat- rics, 27(2), S145–S155.

Eikeseth, S. (2009). Outcome of comprehensive psycho-educational

interventions for young children with autism. Research in Developmental Disabilities, 30, 158–178.

*Eikeseth, S., Smith, T., Jahr, E., & Eldevik, S. (2002). Intensive

behavioral treatment at school for 4- to 7-year-old children with

autism: A 1-year comparison controlled study. Behavior Mod- ification, 26, 49–68.

*Eikeseth, S., Smith, T., Jahr, E., & Eldevik, S. (2007). Outcome for

children with autism who began intensive behavioral treatment

between ages 4 and 7: A comparison study. Behavior Modifi- cation, 31, 264–278.

*Eldevik, S., Eikeseth, S., Jahr, E., & Smith, T. (2006). Effects of

low-intensity behavioral treatment for children with autism and

mental retardation. Journal of Autism and Developmental Disorders, 36, 211–224.

?Eldevik, S., Hastings, R. P., Hughes, C., Jahr, E., Eikeseth, S., &

Cross, S. (2009). Meta-analysis of early intensive behavioral

intervention for children with autism. Journal of Clinical Child and Adolescent Psychology, 38, 439–450.

Eldevik, S., Hastings, R. P., Hughes, J. C., Jahr, E., Eikeseth, S., &

Cross, S. (2010) Using participant data to extend the evidence

base for intensive behavioral intervention for children with

autism. American Journal on Intellectual and Developmental Disabilities, 115, 381–405.

Foxx, R. M. (1993). Sapid effects awaiting independent replication.

American Journal on Mental Retardation, 97, 375–376. Granpeesheh, D., Tarbox, J., & Dixon, D. R. (2009). Applied

behavior analytic interventions for children with autism: A

description and review of treatment research. Annals of Clinical Psychology, 21(3), 162–173.

Green, G., Brennan, L. C., & Fein, D. (2002). Intensive behavioral

treatment for a toddler at high risk for autism. Behavior Modification, 26, 69–102.

Green, V. A., Pituch, K. A., Itchon, J., Choi, A., O’Reilly, M., &

Sigafoos, J. (2006). Internet survey of treatments used by parents

of children with autism. Research in Developmental Disabilities, 27, 70–84.

Gresham, F. M., & MacMillan, D. L. (1998). Early intervention

project: Can its claims be substantiated and its effects replicated?

Journal of Autism and Developmental Disorders, 28, 5–13. *Harris, S. L., & Handleman, J. S. (2000). Age and IQ at intake as

predictors of placement for young children with autism: A four-

to six-year follow-up. Journal of Autism and Developmental Disorders, 30, 137–142.

*Harris, S. L., Handleman, J. S., Gordon, R., Kristoff, B., & Fuentes,

F. (1991). Changes in cognitive and language functioning of

preschool children with autism. Journal of Autism and Devel- opmental Disabilities, 21, 281–290.

Hedges, L., & Olkin, I. (1985). Statistical models for meta-analysis. New York: Academic Press.

Higgins, J. P. T., Deeks, J. J., & Altman, D. G. (2008). Special topics

in statistics. In J. P. T. Higgins & S. Green (Eds.), Cochrane

handbook for systematic reviews of interventions (pp. 481–530). West Sussex, UK: Wiley-Blackwell.

*Howard, J. S., Sparkman, C. R., Cohen, H. G., Green, G., &

Stanislaw, H. (2005). A comparison of intensive behavior

analytic and eclectic treatments for young children with autism.

Research in Developmental Disabilities, 26, 359–383. Kanner, L. (1943). Autistic disturbances of affective contact. Nervous

Child, 2, 217–250. Kratochwill, T. R., & Stoiber, K. C. (2002). Evidence-based

interventions in school psychology: Conceptual foundations of

the procedural and coding manual of division 16 and the society

for the study of school psychology task force. School Psychology Quarterly, 17, 341–389.

Lovaas, O. I. (1981). Teaching developmentally disabled children: The me book. Baltimore: University Park.

*Lovaas, O. I. (1987). Behavioral treatment and normal educational

and intellectual functioning in young autistic children. Journal of Consulting and Clinical Psychology, 55, 3–9.

Lovaas, O. I. (2003). Teaching individuals with developmental delays: Basic intervention techniques. Austin, TX: Pro-Ed.

Love, J. R., Carr, J. E., Almason, S. M., & Petursdottir, A. I. (2009).

Early and intensive behavioral intervention for autism: A survey

of clinical practices. Research in Autism Spectrum Disorders, 3, 421–428.

*Magiati, I., Charman, T., & Howlin, P. (2007). A two-year

prospective follow-up study of community-based early intensive

behavioural intervention and specialist nursery provision for

children with autism spectrum disorder. Journal of Child Psychology and Psychiatry, 48, 803–812.

?Makrygianni, M. K., & Reed, P. (2010). A meta-analytic review of

the effectiveness of behavioural early intervention programs for

children with autism spectrum disorders. Research in Autism Spectrum Disorders, 4, 577–593.

*Matos, M. A., & Mustaca, A. E. (2005). Analisis comportamental

aplicado (ACA y transtornos generalizados del desarrollo

(TGD): Su evaluacion en Argentina [Applied behavior analysis

and pervasive developmental disabilities: Assessment in Argen-

tina]. Interdisciplinaria, 22(1), 59–76. Matson, J. L., & Smith, K. R. M. (2008). Current status of intensive

behavioral interventions for young children with autism and

PDD-NOS. Research in Autism Spectrum Disorders, 2, 60–74. Maurice, C., Green, G., & Foxx, R. M. (Eds.) (2001). Making a

difference: behavioral intervention for autism. Austin, TX: Pro-Ed.

Maurice, C., (Ed.), Green, G., & Luce, S. (Co-Eds.) (1996).

Behavioral intervention for young children with autism: A manual for parents and professionals. Austin, TX: Pro-Ed.

McEachin, J. J., Smith, T., & Lovaas, O. I. (1993). Long-term outcome

for children with autism who received early intensive behavioral

treatment. American Journal on Mental Retardation, 97, 359–372. Mesibov, G. B. (1993). Treatment outcome is encouraging. American

Journal on Mental Retardation, 97, 379–380. Mundy, P. (1993). Normal versus high functioning status in children

with autism. American Journal on Mental Retardation, 97, 381–384.

National Autism Center. (2009). National standards report. Obtained

January 6, 2010 from: www.nationalautismcenter.org/pdf/

NAC%20Standards%20Report.pdf.

Odom, S. L., Brantlinger, E., Gersten, R., Horner, R. H., Thompson,

B., & Harris, K. R. (2005). Research in special education:

Scientifc methods and evidence-based practices. Exceptional Children, 71, 137–148.

*Reed, P., Osborne, L. A., & Corness, M. (2007a). Brief Report:

Relative effectiveness of different home-based behavioural

approaches to early teaching intervention. Journal of Autism and Developmental Disorders, 37, 1815–1821.

123

520 J Autism Dev Disord (2012) 42:512–520

*Reed, P., Osborne, L. A., & Corness, M. (2007b). The real-world

effectiveness of early teaching interventions for children with

autistic spectrum disorders. Exceptional Children, 73, 413–417. Reeves, B. C., Deeks, J. J., Higgins, J. P. T., & Wells, G. A. (2008).

Including non-randomized studies. In J. P. T. Higgins & S. Green

(Eds.), Cochrane handbook for systematic reviews of interventions (pp. 391–432). West Sussex, UK: Wiley-Blackwell.

Reichow, B. (2011). Development, procedures, and application of the

evaluative method for determining evidence-based practices in

autism. In B. Reichow, P. Doehring, D. V. Cicchetti, & F.

R. Volkmar (Eds.), Evidence-based practices and treatments for children with autism (pp. 25–39). New York, NY: Springer.

Reichow, B., Volkmar, F. R., & Cicchetti, D. V. (2008). Development

of an evaluative method for determining the strength of research

evidence in autism. Journal of Autism and Developmental Disorders, 38, 1311–1319.

?Reichow, B., & Wolery, M. (2009). Comprehensive synthesis of

early intensive behavioral interventions for young children with

autism based on the UCLA Young Autism Project model.

Journal of Autism and Developmental Disorders, 39, 23–41. *Remington, B., Hastings, R., Kovshoff, H., degli Espinosa, F., Jahr,

E., Brown, T., et al. (2007). Early intensive behavioral

intervention: Outcomes for children with autism and their

parents after two years. American Journal on Mental Retarda- tion, 112, 418–438.

Rogers, S. J., & Vismara, L. A. (2008). Evidence-based comprehen-

sive treatments for early autism. Journal of Clinical Child and Adolescent Psychology, 37, 8–38.

*Sallows, G. O., & Graupner, T. D. (2005). Intensive behavioral

treatment for children with autism: Four-year outcome and

predictors. American Journal on Mental Retardation, 110, 417–438.

Schopler, E., Short, A., & Mesibov, G. (1989). Relation of behavioral

treatment to ‘‘normal functioning’’: Comment on Lovaas.

Journal of Consulting and Clinical Psychology, 57, 162–164. *Sheinkopf, S. J., & Siegel, B. (1998). Home-based behavioral

treatment of young children with autism. Journal of Autism and Developmental Disorders, 28, 15–23.

Silverman, W. K., & Hinshaw, S. P. (2008). The second special issue

on evidence-based psychosocial treatments for children and

adolescents: A 10-year update. Journal of Clinical Child and Adolescent Psychology, 37, 1–7.

*Smith, T., Eikeseth, S., Klevstrand, M., & Lovaas, O. I. (1997).

Intensive behavioral treatment for preschoolers with severe

mental retardation and 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 intervention for children with pervasive devel-

opmental disorder. American Journal on Mental Retardation, 105, 269–285.

Smith, T., Eikeseth, S., Sallows, G. O., & Graupner, T. D. (2009).

Effcacy of applied behavior analysis in autism. The Journal of Pediatrics, 155, 151–512.

Sparrow, S. S., Balla, D. A., & Cicchetti, D. V. (1984). Vineland Adaptive Behavior Scales. Circle Pines, MN: American Guid- ance Service.

?Spreckley, M., & Boyd, R. (2009). Effcacy of applied behavioral

intervention in preschool children with autism for improving

cognitive, language, and adaptive behavior: A systematic review

and meta-analysis. The Journal of Pediatrics, 154, 338–344. Stahmer, A. C., Collings, N. M., & Palinkas, L. A. (2005). Early

intervention practices for children with autism: Descriptions

from community providers. Focus on Autism and Other Devel- opmental Disabilities, 20, 66–79.

?Virués-Ortega, J. (2010). Applied behavior analytic intervention for

autism in early childhood: Meta-analysis, meta-regression and

dose-response meta-analysis of multiple outcomes. Clinical Psychology Review, 30, 387–399.

*Weiss, M. J. (1999). Differential rates of skill acquisition and

outcomes of early intensive behavioural intervention for autism.

Behavioral Interventions, 14, 3–22. What Works Clearinghouse. (2010). Lovaas model of applied

behavior analysis. Obtained September 9, 2010 from http://ies.

ed.gov/ncee/wwc/pdf/wwc_lovaas_082410.pdf.

123