DEVELOPMENTAL MEDICINE & CHILD NEUROLOGY SYSTEMATIC REVIEW

Do early intervention programmes improve cognitive and motor

outcomes for preterm infants after discharge? A systematic
review
JANE ORTON 1 | ALICIA SPITTLE 1 , 2 , 3 | LEX DOYLE 1 ,2 , 4 | PETER ANDERSON 2 , 5 | ROSLYN BOYD 2 ,6

1 The Royal Women's Hospital, Melbourne, Australia. 2 Murdoch Childrens Research Institute, Melbourne, Australia. 3 School of Physiotherapy, The University of
Melbourne, Melbourne, Australia. 4 Department of Obstetrics and Gynaecology, The University of Melbourne, Melbourne, Australia. 5 School of Behavioural Science,
The University of Melbourne, Melbourne, Australia. 6 Queensland Cerebral Palsy and Rehabilitation Research Centre, University of Queensland, Brisbane, Australia.

Correspondence to Jane Orton at Newborn Research, The Royal Women's Hospital, 20 Flemington Road, Parkville, Victoria 3052, Australia.
E-mail: jane.orton@thewomens.org.au

PUBLICATION DATA
Accepted for publication 1st June 2009.
Published online 1st September 2009.
LIST OF ABBREVIATIONS
PEDro Physiotherapy Evidence Database
SMD Standardized mean difference
TIMP Test of Infant Motor Performance
ACKNOWLEDGEMENTS
Public Health (Dora Lush) NHMRC Postgradu-
ate fellowship (AS); NHMRC Postdoctoral
Training Fellowship (PA), NHMRC Career
Development Fellowship Project Grant (RB):
National Health and Medical Research Council
(NHMRC) 284512. A previous version of this
review has been published in The Cochrane
Library, Issue 2, April 2007.
AIM The aim of this study was to review the effects of early developmental inter-
vention after discharge from hospital on motor and cognitive development in
preterm infants.
METHOD Randomized controlled trials (RCTs) or quasi-RCTs of early develop-
mental intervention programmes for preterm infants in which motor or cognitive
outcomes were reported and in which the intervention commenced before or
after discharge were included. A systematic review and meta-analysis of studies
grouped by intervention, age of outcome, and study quality was undertaken.
Databases searched (up to January 2009) included the Cochrane Central Register
of Controlled Trials, MEDLINE, CINAHL, PsycINFO, and Embase.
RESULTS Eighteen studies met the inclusion criteria (2686 patients randomized),
but only 11 studies had data suitable for meta-analysis. Early developmental
intervention improved cognitive outcomes at infant age (developmental
quotient: standardized mean difference [SMD] 0.42, 95% confidence interval [CI]
0.33–0.52; p<0.001), and at preschool age (IQ: SMD 0.46, 95% CI 0.33–0.59;
p<0.001). However, the benefit was not sustained at school age (IQ: SMD 0.02,
95% CI –0.10 to 0.14; p=0.71). Early intervention had little effect on motor
outcome at infant or school age, and there was no study reporting motor
outcome at preschool age.
INTERPRETATION Current evidence suggests that the benefits of developmental
intervention postdischarge are restricted to short-term gains in cognitive
outcome.

Survival rates in very preterm infants have improved
over the past two decades; however, the rates of neurobeha-
vioural impairments in middle childhood have remained rel-
atively constant, with up to 15% having a diagnosis of
cerebral palsy and 50% exhibiting cognitive, motor, or
behavioural problems.1–3 These neurosensory impairments
are complex and often subtle, and may become more obvious
at school age.4,5 Problems with attention and hyperactivity
are commonly reported,6–8 and minor motor impairments,
classified as developmental coordination disorder, are more
prevalent in children born preterm.9–11 Long-term follow-
up studies indicate that at least some impairments persist in
young adults born at very low birthweight, resulting in a
lower high-school graduation rate compared with those
born at term with normal birthweight.12
Early developmental interventions have been used in the
clinical setting for many years with the aim of improving
the overall outcome of preterm infants. The complex bio-
logical, medical, and environmental elements that contrib-
ute to development have led to programmes that
encompass many different components, with services
provided by a variety of disciplines.13 Different aspects
of early development, such as motor, cognitive, and
behavioural development, as well as mother–infant

ª The Authors. Journal compilation ª Mac Keith Press 2009

DOI: 10.1111/j.1469-8749.2009.03414.x 851
interactions, are emphasized depending on the outcomes
being targeted. However, the effectiveness of these early
developmental intervention programmes in preterm infants
following hospital discharge has not been fully established.
The objective of this study, therefore, was to review sys-
tematically the literature to determine the effects of early
developmental intervention programmes after discharge
from hospital on both the cognitive and motor develop-
ment of preterm infants at infant (0 to <3y), preschool (3 to
<5y), and school age (5 to <13 and 13 to 18y). This paper is
based on a Cochrane review published in The Cochrane
Library 2007, issue 2.14*
METHOD
Literature search strategy
This systematic review followed the guidelines of the Neo-
natal Review Group of the Cochrane Collaboration.15 The
following databases were comprehensively searched by two
reviewers (JO and AS): the Cochrane Central Register of
Controlled Trials (CENTRAL, The Cochrane Library,
issue 4, 2008), MEDLINE Advanced (1966 to January
2009), CINAHL (1982 to January 2009), PsycINFO (1966
to January 2009), and Embase (1988 to January 2009). The
reviewers cross-referenced relevant literature, including
identified trials and existing review articles. The search
strategy comprised the following Medical Subject Head-
ings (MeSH) or keywords: (1) infant, preterm or infant,
low birthweight; (2) and early intervention (education) or
developmental care or physical therapy or occupational
therapy or psychology or parent–infant relationship or reha-
bilitation or exercise or neurodevelopmental therapy or
infant stimulation; (3) and child development or infant
development or cognition or intellectual disability or devel-
opmental disabilities or psychomotor performance or psy-
chomotor disorders or cerebral palsy or developmental
coordination disorder or movement disorders or motor skill
disorders; (4) not drug therapy or genetics or chest physio-
therapy or cardiac.
Studies that are reported in English or a language for
which a translator was available were included.
Selection criteria
Studies had to meet the following selection criteria
for inclusion: (1) randomized controlled trial (RCT) or
quasi-RCT; (2) participants comprised infants born at <37
weeks’ gestational age with no major congenital abnormal-
ities; (3) intervention commenced while the infant was still
in hospital or after discharge but at age less than 12
*Cochrane reviews are regularly updated as new evidence emerges and
in response to feedback, and The Cochrane Library should be con-
sulted for the most recent version of the review.
852 Developmental Medicine & Child Neurology 2009, 51: 851–859
months; (4) early intervention that aimed to improve cog-
nitive or motor outcomes, performed at home, hospital, or
in a community centre; (5) programme focus was on the
parent–infant relationship, infant development, or both; (6)
interventions included physical therapy, occupational ther-
apy, psychology, neurodevelopmental therapy, parent–
infant relationship enhancement, infant stimulation, infant
development, developmental care, or early intervention
(education); (7) outcome assessed using motor and ⁄ or cog-
nitive measures.
Validity assessment
The methodological quality of included trials was assessed
by three independent raters using the Physiotherapy Evi-
dence Database (PEDro) scale16 adapted from the Delphi
list. The PEDro scale includes 11 criteria to assess the
internal and external validity of RCTs. However, the maxi-
mum possible score on the PEDro scale for this type of
study was 10. One of the criteria is blinding of therapists,
which is not possible when providing intervention of this
type. After discussion there was no disagreement between
raters on the PEDro scores.
Data abstraction
The information extracted from the studies included study
design and population demographics (Table I). The theo-
retical structure, content, and dosage of the intervention
programmes were tabulated (Table SI, supporting infor-
mation published online). For individual trials, where pos-
sible, mean values for treatment and comparison groups
(and 95% confidence intervals [CIs]) were reported for
continuous variables (Tables SII and SIII supporting infor-
mation published online). Study authors were contacted
for missing data.
Quantitative data synthesis
RevMan 5.0.16 (The Cochrane Collaboration, 2008,
Copenhagen, Denmark) was used for data management
and analysis with data independently entered by two
reviewers. For data analysis, ‘treatment group’ refers to
infants who were involved in early developmental inter-
vention programmes and ‘comparison group’ refers to
infants who received standard medical follow-up. For con-
tinuous outcomes, standardized mean differences (SMDs)
were calculated as there were a variety of outcome mea-
sures (with different standard deviations). Cognitive and
motor outcome data were pooled into four age bands:
infant (0 to <3y), preschool (3 to <5y), or school ages (5
to <13 and 13 to 18y). If studies reported data at more
than one time point within an age band, data from the lat-
est assessment were used. Pooled effects for treatment
effect were calculated across the trials using a fixed-effects
 Table I: Characteristics and methods of studies included in analysis

Group allocation
Individuals PEDro No. of GA Birthweight
Study enrolled (n) Method score sites (wks) (g) Treatment (n) Comparison (n)

Rice31 30 RCTa 6 1 <37 NS 15 15

27
Field et al. 60 RCTa 6 NS <37 <2500 30 30
Nurcombe et al.,23 78 RCT 8 1 <37 <2250 38 40
Achenbach et al.,39,40
Rauh et al.41
Goodman et al.,25 107 Quasi-RCT 6 1 <34 1700 40b 40b
42
Rothberg et al.
Barrera et al.,19 80 RCTa 5 3 £37 <2000 38b 21b
43
Barerra et al.,
Barerra and Kitching,44
Barerra et al.45
Piper et al.29 134 RCT 6 2 NS <1500c 56b 59b
30
Resnick et al. 41 Quasi-RCT 5 1 NS <1800 21 20
IHDP,20 Blair et al.,46 985 RCT 9 8 <37 <2500 377 608
Berlin et al.,47 Blair,48
Brooks-Gunn et al.,49–51
McCormick et al.,52–54
APIP,17 Johnson et al.38 284 RCT 9 2 <33 NS 187 97
Bao et al.18 103 Quasi-RCT 5 3 >28–<37 NS 52 51
Lekskulchai and Cole28 84 RCT 8 1 <37d NS 43 41
Melnyk et al.21 55 Quasi-RCT 7 1 26–36 <2500 26 29
Nelson et al.22 37 RCTa 6 1 23–36 <1500d 21 16
Yigit et al.32 190 RCTa 5 1 <34 <2000 80b 80b
Ohgi et al.24 24 RCT 8 1 NS <2500d 12 12
Cameron et al.26 72 RCT 7 1 <32 <1500 34 38
Kaaresen et al.33 146 RCT 9 1 NS <2000 72 74
Koldewijn et al.34 176 RCT 8 7 <32 <1500 86 90

a
Method of randomization unclear. bNumber of individuals in group at follow-up. It is, however, unclear how many individuals were
initially randomized to treatment and comparison groups. cIncludes a subgroup of infants who weighed >1500g at birth and who
experienced birth asphyxia, seizures, or central nervous system dysfunction. dInfants were considered at high risk of neurological
sequelae owing to periventricular leukomalacia or abnormal ultrasound findings, or based on assessment at term. PEDro, Physiotherapy
Evidence Database; GA, gestational age; RCT, randomized controlled trial; NS, not specified.

model. To guard against an over-influence of small-study RESULTS

effects, we also used a random-effects model, in which
heterogeneity was significant. As no statistical conclusion
was altered, we report only results from fixed-effects mod-
els. A sensitivity analysis was performed using the rating
of study quality (PEDro scale) to assess the effect of bias
on results and to reduce heterogeneity. RCTs that scored
7 or more were considered to have strong validity and
were considered high-quality studies; the remainder were
lower-quality studies. We plotted the intervention effect
estimates from individual studies against the study’s stan-
dard error, known as a ‘funnel plot’, to assess the possibil-
ity of publication bias for both cognitive and motor
outcomes (Fig. 2).
Details of the screening process to identify the trials in the
final meta-analysis (trial flow) are shown in Fig. 1. There were
18 RCTs or quasi-RCTs of early developmental interven-
tions after hospital discharge17–34 involving 2686 infants
(Table I). Eleven of these had data suitable for use in meta-
analysis (Tables SII and SIII).17–25,33,34 The structure and
content of the intervention programmes are detailed in
Table SI.
Qualitative analysis
Nine of the 18 studies had a PEDro score of 7 or more
and were considered of high quality (internal validity;
Table I). However, one of these studies used quasi-

Early Intervention for Preterm Infants Jane Orton et al. 853

 opmental intervention (n=825) scored a mean developmen-
Table II: Meta-analysis: effect of early developmental intervention on
tal quotient of 0.42SD (p<0.001) higher than infants who
cognitive outcome
received standard follow-up (n=924). Six of these studies
SMD (95% CI) fixed
were classified as having high-quality methodology. Sensi-
tivity analysis of the high-quality studies showed a signifi-
Infant age (0<3y) cant treatment effect supporting the intervention
Higher-quality studies (developmental quotient; SMD 0.40; p<0.001). There was
Nurcombe et al.23 0.50 ()0.04 to 1.05) significant heterogeneity among the studies overall
IHDP20 0.51 (0.37 to 0.65)
(Table II). After sensitivity analysis, the significant hetero-
APIP17 0.19 ()0.06 to 0.45)
geneity persisted only in the lower-quality studies group.
Ohgi et al.24 0.54 ()0.30 to 1.37)
There was insufficient statistical support for a differential
Kaaresen et al.33 0.22 ()0.12 to 0.56)
treatment-group effect across the two subgroups (v2 test
Koldewijin et al.34 0.25 ()0.05 to 0.55)
Subtotal 0.40 (0.29 to 0.50); I 2=31%; z=7.2, p<0.001
for subgroup differences 1.64; p=0.20).
Lower-quality studies At preschool age (3 to <5y), three studies reported cog-
Goodman et al.25 0.05 ()0.39 to 0.48) nitive outcome, and these data were pooled for meta-analy-
Bao et al.18 1.26 (0.77 to 1.75) sis. The mean IQ of children who received early
Melnyk et al.21 0.73 (0.11 to 1.36) developmental intervention as infants (n=403) was 0.42SD
Nelson et al.22 0.39 ()0.38 to 1.16) (p<0.001) higher than that of children who received stan-
Subtotal 0.59 (0.32 to 0.86); I 2=78%; z=4.24, p<0.001 dard follow-up (n=603). Sensitivity analysis of the high-
Total 0.42 (0.33 to 0.52); I 2=60%; z=8.52, p<0.001 quality studies (n=2) showed a significant treatment effect
Preschool age (3 to <5y)
supporting the intervention (IQ: SMD 0.48SD; p<0.001).
Higher-quality studies
There was no significant heterogeneity.
Nurcombe et al.23 0.79 (0.23 to 1.35)
At school age (5 to <13y), four studies reported cogni-
IHDP20 0.46 (0.33 to 0.60)
tive outcome. Meta-analysis was used for the three studies
Subtotal 0.48 (0.35 to 0.61); I 2=18%; z=7.13, p<0.001
Lower-quality studies
that assessed IQ, all of which were classified as high qual-
Barrera et al.45 )0.05 ()0.68 to 0.58); I 2=NA; z=0.14, p=0.89 ity. School-age children who received early developmental
Total 0.46 (0.33 to 0.59); I 2=47%; z=6.95, p<0.001 intervention (n=484) did not score higher than the
School age (5–13y) children who received standard follow-up (n=627) on IQ
Higher-quality studies measures (IQ: SMD 0.02; p=0.71). There was significant
Nurcombe et al.23 1.02 (0.45 to 1.59) heterogeneity but there were too few studies for a
IHDP20 )0.01 ()0.15 to 0.13) sensitivity analysis.
APIP17 )0.09 ()0.40 to 0.21) At adolescence (age 13–18y), one high-quality study, the
Total 0.02 ()0.10 to 0.14); I 2=84%; z=0.37, p=0.71
Infant Health and Development Program, reported cogni-
SMD, standardized mean difference; CI, confidence interval; I 2, test
tive outcomes and found little evidence to suggest that the
statistic for heterogeneity; z, test statistic for overall effect; IDHP, intervention had a beneficial effect on cognitive develop-
Infant Health and Development Programme; APIP, Avon ment.
Premature Infant Project.
Motor outcomes
randomization and therefore was not included as a high- Seventeen studies reported on motor outcomes, of which
quality study in the sensitivity analysis.21 Four studies eight provided sufficient data for meta-analysis at infant
achieved a PEDro score of only 5. All studies had similar age18,20,22–25,33,34: seven with the Bayley Scales of Infant
prognostic characteristics of intervention and comparisons Development Psychomotor Developmental Index (2nd
at baseline related to perinatal factors such as gestational edition) and one with the Griffiths Mental Development
age; however, two studies reported differences in sociode- Scale Locomotor Sub-scale (Table SIII). At infant age
mographic variables.17,23 there was no difference in motor outcome between infants
who received early developmental intervention (n=632)
Cognitive outcomes and those who received standard follow-up (n=822; SMD
Fifteen studies reported cognitive outcomes at different 0.07; p=0.18). Five of the eight studies at infant age were
ages, of which 11 provided sufficient data for meta-analysis considered to be of high quality.20,23,24,33,34 Sensitivity
(Table SII).17–25,33,34 At infant age (0 to <3y) 10 studies analysis showed no intervention effect (SMD 0.04; p=0.47).
reported sufficient data (means and SDs) to be pooled for Of the studies that did not provide data suitable for meta-
meta-analysis (Table II). Infants who received early devel- analysis at infant age, one study reported a significant dif-

854 Developmental Medicine & Child Neurology 2009, 51: 851–859

 Potentially relevant RCTs or quasi-RCTs
identified and screened by title and abstract.
(n=1129)

RCTs excluded (n=1067)

RCTs and quasi-RCTs retrieved for more
detailed examination by two independent
reviewers (n=62)
Inappropriate intervention, no intervention,
inappropriate population or outcomes,
Not RCT or quasi-RCT

RCTs excluded (n=22)

Not RCT or quasi-RCT, no intervention,
not preterm infants

RCTs and quasi-RCTs included in the review

(n=40. This comprised 18 different studies
(Table I). These were again reviewed to
determine which could be used in meta-analysis
RCTs and quasi-RCTs excluded from meta-
analysis (n=7)
No data available, outcome measure unable to
be pooled with others, number of participants at
time of assessment unclear.

RCTs with usable information by outcome and

included in meta-analysis (n=11)

Figure 1: Included and excluded studies.

ference at infant age.19,27,29–32,34 There was no significant

Table III: Meta-analysis: effect of early developmental intervention heterogeneity among studies (Table III).
on motor outcome
Two studies that reported motor outcomes at school
age (6–12y) but with data not suitable for pooling.17,25
SMD (95% CI) fixed
Goodman et al.25 demonstrated no difference in motor
Infant age (0–2y) outcomes between intervention and comparison groups
Higher-quality studies measured by the Griffiths Mental Development Scale
Nurcombe et al.23 )0.30 ()0.77 to 0.17) Locomotor Sub-scale. The study by the Avon Preterm
IHDP20 0.03 ()0.11 to 0.16) Infant Project17 also found no difference in the rate of
Ohgi et al.24 0.39 ()0.44 to 1.21) motor impairment between children who had received
Kaaresen et al.33 0.07 ()0.27 to 0.41) developmental intervention and those who had standard
Koldewijn et al.34 0.19 ()0.12 to 0.49)
follow-up. There was no study with follow-up of motor
Subtotal 0.04 ()0.07 to 0.15); I 2=0%; z=0.73, p=0.47
outcomes for preterm infants at adolescent age.
Lower-quality studies
A funnel plot of the cognitive studies was reasonably
Goodman et al.25 0.18 ()0.26 to 0.62)
Bao et al.18 0.35 ()0.11 to 0.80)
symmetrical, and thus revealed little evidence of publica-
Nelson et al.22 0.32 ()0.44 to 1.09)
tion bias (Fig. 2a). The funnel plot for motor outcomes
Subtotal 0.27 ()0.02 to 0.56); I 2=0%; z=1.82, p=0.07 was less symmetrical but, partly because the number of
Total 0.07 ()0.03 to 0.18); I 2=0%; z=1.34, p=0.18 studies was lower, again the evidence for publication bias
was not strong (Fig. 2b). There were too few studies for
SMD, standardized mean difference; CI, confidence interval; other ages to allow for meaningful funnel plots.
IDHP, Infant Health and Development Programme; I 2, test statistic
for heterogeneity; z, test statistic for overall effect.
DISCUSSION
The primary objective of this systematic review was to
ference in favour of the intervention group28 (p<0.001) assess the effectiveness of early developmental intervention
using the Test of Infant Motor Performance (TIMP) at 4 programmes after discharge from hospital on cognitive
months of age. The remaining seven studies found no dif- and motor development in preterm infants at infant (0 to

Early Intervention for Preterm Infants Jane Orton et al. 855

a Cognitive outcomes or neurobehavioural disorders as a result of preterm birth.
SE(SMD) Meta-analysis demonstrated a treatment effect of early
0
intervention on cognitive outcomes at infant and preschool
0.1
age of approximately half a standard deviation, an effect
size considered to be clinically important.2 However, at
school age no substantial treatment effect on cognitive out-
0.2
comes was observed. Meta-analysis indicated that early
developmental interventions had no beneficial impact on
0.3
motor outcomes at infant or school age. Significant hetero-
geneity was observed between the studies in this meta-
0.4 analysis with regard to the primary outcome measures, the
focus of the intervention, and the method of delivery. For
SMD example, the primary outcome in some trials was motor
0.5
–1 –0.5 0 0.5 1
Subgroups development, whereas in others it was cognitive develop-
Higher-quality studies ment or both. Further, some interventions focused on the
Lower-quality studies
parent–infant relationship, whereas other interventions
b Motor outcomes
SE(SMD)
focused on the infants’ development or both.
0
The methodological quality of the studies in this review
was variable, with only 9 of the 18 studies considered to be
0.1 of high quality. In general, statistical conclusions were
unaffected when the meta-analysis was restricted to high-
0.2 quality studies. A limitation with intervention trials of
developmental interventions is that it is not feasible to
0.3 mask the person implementing the intervention or the reci-
pient of the intervention (in this case the mother and
0.4 infant) unless the study includes a comparison group
receiving an alternative intervention rather than no treat-
0.5 SMD ment. Only one study in the meta-analysis included a treat-
–1 –0.5 0 0.5 1
Subgroups
ment comparison group rather than a nontreatment
Higher-quality studies comparison group21; however, this was not a randomized
Lower-quality studies
trial. The Avon Premature Infant Project study17 was the
only study to attempt to control for the parent support
Figure 2: Funnel plots for cognitive (a) and motor outcomes (b) at component of early intervention by having three groups:
infant age. The individual study's standard error (SE[SMD]) is plotted one that received an infant development programme, one
against the standardised mean difference (SMD) for the study. The that received parent support only, and one that received
vertical line is the overall SMD. Higher- and lower-quality studies are standard medical care. It should be noted that the assess-
indicated separately, as shown.
ment of methodological quality used in this review (PEDro
scale) did not take into account sample size, which was
quite small for many of the included studies.
<3y), preschool (3 to <5y), and school ages (5 to <13 and A potential limitation of RCTs of developmental inter-
13 to 18y). To our knowledge, this is the first reported ventions is contamination, as families of preterm infants
meta-analysis of early developmental interventions postdis- may seek additional treatments if their child is perceived to
charge from hospital for preterm infants. A recent review35 be ‘at risk’ of developmental difficulties. One study did
explored the effects on motor outcomes of a range of early indicate that referrals to physiotherapy or other specialists
interventions that commenced from birth to 18 months in could occur at any time during follow-up but did not pro-
infants at risk for, or with, developmental motor disorders. vide details.29 In this case, the population included a group
In the Neonatal Individualised Developmental Care Pro- that was identified as having an established impairment,
gram, interventions that focus on specific motor training and intervention received by the infants was aimed at mini-
were found to have a positive effect on motor outcomes in mizing disability. One study identified infants at risk of
high-risk infants.35 In contrast, this systematic review motor impairment based on their TIMP score, and inter-
focused on interventions targeted at preterm infants, a spe- vention was again targeted to the impairment.28 This study
cific population at variable risk of developmental problems found a significant effect of intervention on motor out-

856 Developmental Medicine & Child Neurology 2009, 51: 851–859

come at 4 months. It may be that targeting intervention
and using outcome measures more specific to the out-
comes is a more responsive way to assess intervention. The
populations in the other studies consisted of preterm
infants at variable biological risk but without an identified
impairment. No study identified populations in terms of
the infants’ environmental risk.
The majority of included studies involved infants born
in the 1980s, and the gestational age at entry tended to be
higher than in more recent studies involving infants born
in the 1990s. Perinatal care obviously differs across studies,
and pooling results across eras may result in comparing
different groups of infants with respect to outcomes.1 As
more recent studies are published it may be possible to
group studies according to their era of perinatal care.
Retention of participants was extremely variable
between studies, ranging from 46 to 97%. Low retention
rates can lead to biased outcome assessments. Retention
may be affected also by the burden associated with the
intervention for families, such as accessibility (home or
centre based), duration of sessions, the demands on partici-
pants (parent and infant), time and resources, as well as
satisfaction with the programme. In general, as the time
period between intervention cessation and follow-up
increased, retention rates declined. Although the Infant
Health and Development Program had a demanding pro-
gramme (3y of intervention), it was able to achieve reten-
tion rates of 88% and 65% at 8 and 18 years
respectively,36,37 making it a major contributor. In addition
to reduced retention rates, analysis of the long-term effects
of early developmental interventions was further limited
by the small number of studies that included school-age
follow-up. The amount of intervention received by infants
and families varied greatly between studies (4–336 sessions)
and within studies (e.g. compliance), and it may be expected
that dose and compliance are associated with the effective-
ness of the intervention.
The early developmental interventions included in this
review varied in theoretical content, environmental con-
text, intensity, and length of follow-up and follow-up rates.
This resulted in significant levels of heterogeneity when
pooling outcomes. As a result, assessing which components
and methods of intervention are most beneficial is difficult.
Moreover, we could not devise a system for classifying the
motor interventions. This is further compounded by the
wide variety of measurement tools used in the studies, par-
ticularly for motor assessment. The effect of early interven-
tion on motor development could be analysed only at
infant age using the Bayley Scales of Infant Development
Psychomotor Developmental Index or the Griffiths Mental
Development Scale Locomotor Sub-scale, and there was
no treatment effect of intervention with either assessment.
These are both broad measures of motor development and
do not specifically evaluate minor motor problems. This
makes it difficult to distinguish whether there is an effect
on motor outcome and, if not, whether the measures are
sensitive enough to detect the effects of intervention. Spe-
cific motor assessments, such as the Alberta Infant Motor
Scale, the TIMP, and the Movement Assessment Battery
for Children were used only by individual studies,26,28,38
and these data could not be pooled for meta-analysis. Lek-
skulchai and Cole28 demonstrated a significant effect of
intervention using the TIMP with short-term (4mo cor-
rected age) follow-up. The diversity of motor assessment
tools and the lack of data at older ages limit the ability
to compare results between studies and to make judge-
ments about the effect of early intervention on motor
outcomes.
The age at assessment of cognitive and motor outcomes
may also be an important factor in assessing intervention
effectiveness. Interestingly, only 4 of the 10 studies that
reported on cognitive outcome up to 1 year showed a
significant effect favouring the intervention group. How-
ever, at 2 years, two studies that had previously shown no
difference at 1 year between treatment and comparison did
detect a significant effect favouring the intervention
group.20,23 This may reflect the reliability and validity of
the outcome measures at 1 year compared with 2 years,
or perhaps the effect of intervention is more apparent
at 2 years.
There are limitations to meta-analysis, especially if the
outcomes are too diverse, if the individual studies are at
risk of bias, or if there is evidence of publication bias. We
have attempted to minimize these problems by comparing
similar outcomes within narrow age ranges, within the lim-
itations outlined previously, by assessing the quality of the
studies and highlighting the higher-quality studies, and by
demonstrating that there was little evidence of publication
bias.
CONCLUSION
Early developmental interventions have a positive short-
term effect on the cognitive development of preterm
infants, but these benefits are not seen at school age.
Intervention has a limited effect of on pooled motor out-
comes at any age; however, the paucity of motor assess-
ment tools limits the conclusions that can be drawn. This
review has highlighted the variability in early developmen-
tal intervention programmes and the difficulty in identify-
ing the most effective intervention package. Given the
resources required to provide early developmental inter-
ventions, there is a particular need for further research on
the effects of early developmental intervention pro-
grammes throughout childhood using outcome measures
Early Intervention for Preterm Infants Jane Orton et al. 857
that are relevant to the intervention and sensitive enough
to detect subtle changes. It is important to identify the
target population, the aspects of programmes that are par-
ticularly useful, and the infants and families that benefit
most from such programmes, to target intervention more
effectively. Longer-term benefits to the child and family
must also be considered when providing intervention.
Future studies may include measures that assess not only
functional outcomes, such as educational difficulties,
behavioural, and social problems, but also the impact on
societal participation.
SUPPORTING INFORMATION
Additional supporting information may be found in the
online version of this article:
Table SI: Structure and content of intervention pro-
grammes.
Table SII: Summary of results of studies reporting on
cognitive outcome.
Table SIII: Summary of results of studies reporting on
motor outcome.
This material is available as part of the online article from
http://dx.doi.org/10.1111/j.1469-8749.2009.03414.x (this
will link you directly to the article).
Please note: Wiley-Blackwell are not responsible for the
content or functionality of any supporting materials
supplied by the authors. Any queries (other than missing
material) should be directed to the corresponding author
for the article.

REFERENCES birthweight preschoolers. Dev Med Child 16. Verhagen AP. The Delphi list: a criteria
1. Doyle LW. Evaluation of neonatal inten- Neurol 1996; 38: 927–40. list for quality assessment of randomised
sive care for extremely low birth weight 9. Marlow N, Roberts BL, Cooke RW. clinical trials for conducting systematic
infants in Victoria over two decades. I. Motor skills in extremely low birthweight reviews. J Clin Epidemiol 1998; 51: 1235–
Effectiveness. Pediatrics 2004; 113: 505–09. children at the age of 6 years. Arch Dis 41.
2. Bhutta AT, Cleves MA, Casey PH, Crad- Child 1989; 64: 839–47. 17. APIP; Avon premature Infant Project.
dock MM, Anand KJS. Cognitive and 10. Holsti L, Grunau RV, Whitfield MF. Randomised trial of parental support for
behavioral outcomes of school-age chil- Developmental coordination disorder in families with very preterm children. Arch
dren who were born preterm: a meta-anal- extremely low birth weight children at Dis Child Fetal Neonatal Ed 1998; 79: F4–
ysis. JAMA 2002; 288: 728–37. nine years. J Dev Behav Pediatr 2002; 23: 11.
3. Pederson SJ, Sommerfelt K, Missiuna C. 9–15. 18. Bao X, Sun S, Wei S. Early intervention
An international consensus on children 11. Powls A, Botting N, Cooke RW, Marlow promotes intellectual development of pre-
with developmental coordination disor- N. Motor impairment in children 12 to mature infants: a preliminary report. Early
ders. Can J Occup Ther 1995; 62: 3–6. 13 years old with a birthweight of less than Intervention of Premature Infants Cooper-
4. Anderson P, Doyle LW, Callanan C, et al. 1250g. Arch Dis Child Fetal Neonatal Ed ative Research Group. Chin Med J (Engl)
Neurobehavioral outcomes of school-age 1995; 73: F62–6. 1999; 112: 520–23.
children born extremely low birth weight 12. Hack M, Flannery D, Schluchter M, et al. 19. Barrera M, Cunningham C, Rosenbaum
or very preterm in the 1990s. JAMA 2003; Outcomes in young adulthood for very- P. Low birth weight and home interven-
289: 3264–72. low-birth-weight infants. N Engl J Med tion strategies: preterm infants. J Dev
5. Hall A, McLeod A, Counsell C, Thomson 2002; 346: 149–57. Behav Pediatr 1986; 7: 361–66.
L, Mutch L. School attainment, cognitive 13. Berger SE, Holt-Turner I, Cupoli JM, 20. IHDP. Enhancing the outcomes of low-
ability and motor function in a total Scot- Mass M, Hageman JR. Caring for the birth-weight, premature infants. A multi-
tish very-low-birthweight population at graduate from the neonatal intensive care site, randomized trial. The Infant Health
eight years: a controlled study. Dev Med unit: at home, in the office, and in the and Development Program. JAMA 1990;
Child Neurol 1995; 37: 1037–50. community. Pediatr Clin North Am 1998; 263: 3035–42.
6. Botting N, Powls A, Cooke RWI, Marlow 45: 701–12. 21. Melnyk BM, Alpert-Gillis L, Feinstein
N. Cognitive and educational outcome of 14. Spittle AJ, Orton J, Doyle LW, Boyd R. NF, et al. Improving cognitive develop-
very-low-birthweight children in early Early developmental intervention pro- ment of low-birth-weight premature
adolescence. Dev Med Child Neurol 1998; grams post hospital discharge to prevent infants with the COPE program: a pilot
40: 652–60. motor and cognitive impairments in pre- study of the benefit of early NICU inter-
7. Hoy EA, Sykes DH, Bill JM, Halliday term infants. Cochrane Database Syst Rev vention with mothers. Res Nurs Health
HL, McClure BG, Reid MM. The social 2007; Issue 2: CD005495. 2001; 24: 373–89.
competence of very-low-birthweight chil- 15. Higgins JPT, Green S. Cochrane hand- 22. Nelson MN, White-Traut RC, Vasan U,
dren: teacher, peer, and self-perceptions. book for systematic reviews of intervention et al. One-year outcome of auditory-tac-
J Abnorm Child Psychol 1992; 20: 123–50. 4.2.5. In: Higgins JPT, Green S, editors. tile-visual-vestibular intervention in the
8. Sommerfelt K, Troland K, Ellertsen B, The Cochrane Library. Chichester: John neonatal intensive care unit: effects of
Markstad T. Behavioral problems in low- Wiley and Sons Ltd, 2005. severe prematurity and central nervous

858 Developmental Medicine & Child Neurology 2009, 51: 851–859

 system injury. J Child Neurol 2001; 16: children: outcome at 2 years. Early Hum Topics in Early Childhood Special Education
493–98. Dev 2008; 84: 201–09. 1991; 10: 14–23.
23. Nurcombe B, Howell DC, Rauh VA, Teti 34. Koldewijn K, Wolf M-J, van Wassnaer A, 45. Barrera ME, Rosenbaum PL, Cunning-
DM, Ruoff P, Brennan J. An intervention et al. The Infant Behavioral Assessment ham CE. Early home intervention with
program for mothers of low-birthweight and Intervention Program for very low low-birth-weight infants and their parents.
infants: preliminary results. J Am Acad birth weight infants at 6 months corrected Child Dev 1986; 57: 20–33.
Child Psychiatry 1984; 23: 319–25. age. J Pediatr 2009; 154: 33–8. 46. Blair C, Ramey CT, Hardin JM. Early
24. Ohgi S, Fukuda M, Akiyama T, Gima H. 35. Blauw-Hospers CH, Hadders-Algra M. A intervention for low birthweight, pre-
Effect of an early intervention programme systematic review of the effects of early mature infants: participation and intellec-
on low birthweight infants with cerebral intervention on motor development. Dev tual development. Am J Ment Retard
injuries. J Paediatr Child Health 2004; 40: Med Child Neurol 2005; 47: 421–32. 1995; 99: 542–54.
689–95. 36. McCarton CM, Brooks-Gunn J, Wallace 47. Berlin LJ, Brooks-Gunn J, McCarton C,
25. Goodman M, Rothberg AD, Houston- IF, et al. Results at age 8 years of early McCormick MC. The effectiveness of
McMillan JE, Cooper PA, Cartwright JD, intervention for low-birth-weight pre- early intervention: examining risk factors
van der Velde MA. Effect of early neuro- mature infants: the Infant Health and and pathways to enhanced development.
developmental therapy in normal and at- Development Program. JAMA 1997; 277: Prev Med 1998; 27: 238–45.
risk survivors of neonatal intensive care. 126–32. 48. Blair C. Early intervention for low birth
Lancet 1985; ii: 1327–30. 37. McCormick MC, Brooks-Gunn J, Buka weight, preterm infants: the role of nega-
26. Cameron EC, Maehle V, Reid J. The SL, et al. Early intervention in low birth tive emotionality in the specification of
effects of an early physical therapy inter- weight premature infants: results at effects. Dev Psychopathol 2002; 14: 311–
vention for very preterm, very low birth 18 years of age for the Infant Health and 32.
weight infants: a randomized controlled Development Program. Pediatrics 2006; 49. Brooks-Gunn J, Klebanov P, Liaw FR,
clinical trial. Pediatr Phys Ther 2005; 17: 117: 771–80. Spiker D. Enhancing the development
107–19. 38. Johnson S, Ring W, Anderson P, Marlow of low-birthweight, premature infants:
27. Field TM, Widmayer SM, Stringer S, Ig- N. Randomised trial of parental support changes in cognition and behavior over the
natoff E. Teenage, lower-class, black for families with very preterm children: first three years. Child Dev 1993; 64: 736–
mothers and their preterm infants: an outcome at 5 years. Arch Dis Child 2005; 53.
intervention and developmental follow-up. 90: 909–15. 50. Brooks-Gunn J, Liaw FR, Kato Kleba-
Child Dev 1980; 51: 426–36. 39. Achenbach TM, Howell CT, Aoki MF, nov P. Effects of early intervention on
28. Lekskulchai R, Cole J. Effect of a develop- Rauh VA. Nine-year outcome of the cognitive function of low birth weight
mental program on motor performance in Vermont intervention program for low preterm infants. J Pediatr 1992; 120:
infants born preterm. Aust J Physiother birth weight infants. Pediatrics 1993; 91: 350–59.
2001; 47: 169–76. 45–55. 51. Brooks-Gunn J, McCarton CM, Casey
29. Piper MC, Kunos VI, Willis DM, Mazer 40. Achenbach TM, Phares V, Howell CT, PH, et al. Early intervention in low-birth-
BL, Ramsay M, Silver KM. Early physical Rauh VA, Nurcombe B. Seven-year out- weight premature infants: results through
therapy effects on the high-risk infant: a come of the Vermont intervention pro- age 5 years from the Infant Health and
randomized controlled trial. Pediatrics gram for low birth weight infants. Child Development Program. JAMA 1994; 272:
1986; 78: 216–24. Dev 1990; 61: 1672–81. 1257–62.
30. Resnick MB, Armstrong S, Carter RL. 41. Rauh VA, Achenbach TM, Nurcombe B, 52. McCormick MC, Brooks-Gunn J, Work-
Developmental intervention program for Howell CT, Teti DM. Minimizing man-Daniels K, Turner J, Peckham GJ.
high-risk premature infants: effects on adverse effects of low birthweight: four- The health and developmental status of
development and parent–infant interac- year results of an early intervention pro- very low-birth-weight children at school
tions. J Dev Behav Pediatr 1988; 9: 73–8. gram. Child Dev 1988; 59: 544–53. age. JAMA 1992; 267: 2204–08.
31. Rice RD. The effects of the rice infant 42. Rothberg AD, Goodman M, Jacklin LA, 53. McCormick MC, McCarton C, Brooks-
sensorimotor stimulation treatment on the Cooper PA. Six-year follow-up of early Gunn J, Belt P, Gross RT. The Infant
development of high-risk infants. Birth physiotherapy intervention in very low Health and Development Program:
Defects Orig Artic Ser 1979; 15: 7–26. birth weight infants. Pediatrics 1991; 88: interim summary. J Dev Behav Pediatr
32. Yigit S, Kerem M, Livanelioglu A, et al. 547–52. 1998; 19: 359–70.
Early physiotherapy intervention in pre- 43. Barrera M, Doucet D, Kitching K. Early 54. McCormick MC, McCarton C, Tonascia
mature infants. Turk J Pediatr 2002; 44: home intervention and socio-emotional J, Brooks-Gunn J. Early educational inter-
224–29. development of preterm infants. Infant vention for very low birth weight infants:
33. Kaaresen PI, Ronning JA, Tunby J, Nord- Ment Health J 1990; 11: 142–57. results from the Infant Health and Devel-
hov SM, Ulvund SE, Dahl LB. A random- 44. Barrera M, Kitching K. A 3-year early opment Program. J Pediatr 1993; 123:
ized controlled trial of an early home intervention follow-up study with 527–33.
intervention program in low birth weight low birth weight infants and their parents.

Early Intervention for Preterm Infants Jane Orton et al. 859