Early Human Development 86 (2010) 413–417

Contents lists available at ScienceDirect

Early Human Development

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e a r l h u m d ev

Predictability of cerebral palsy in a high-risk NICU population

E. Himpens a,⁎, A. Oostra b, I. Franki a, S. Vansteelandt c, P. Vanhaesebrouck d, C. Van den Broeck a
a
Rehabilitation Sciences and Physiotherapy Ghent, University College Arteveldehogeschool–Ghent University, Ghent, Belgium
b
Centre for Developmental Disorders, Ghent, Belgium
c
Department of Applied Mathematics and Computer Science, Ghent University, Ghent, Belgium
d
Department of Neonatology, University Hospital Ghent, Ghent, Belgium

a r t i c l e i n f o a b s t r a c t

Article history: Aim: This study aims to create a predictive model for the assessment of the individual risk of developing
Received 30 March 2010 cerebral palsy in a large cohort of selected high-risk infants.
Received in revised form 17 May 2010 Patients and methods: 1099 NICU-admitted high-risk infants were assessed up to the corrected age of at least
Accepted 18 May 2010 12 months. CP was categorized relative to subtype, distribution and severity. Several perinatal characteristics
(gender, gestational age, multiple gestation, small for gestational age, perinatal asphyxia and duration of
Keywords:
mechanical ventilation), besides neonatal cerebral ultrasound data were used in the logistic regression
Cerebral palsy
model for the risk of CP.
Characteristics of CP
Risk factors for CP
Results: Perinatal asphyxia, mechanical ventilation N 7 days, white matter disease except for transient
Prediction of CP echodensities b 7 days, intraventricular haemorrhage grades III and IV, cerebral infarction and deep grey
matter lesions were recognized as independent predictors for the development of CP. 95% of all children with
CP were correctly identified at or above the cut-off value of 4.5% probability of CP development. Higher
gestational age, perinatal asphyxia and deep grey matter lesion are independent predictors for non-spastic
versus spastic CP (OR = 1.1, 3.6, and 7.5, respectively). Independent risk factors for prediction of unilateral
versus bilateral spastic CP are higher gestational age, cerebral infarction and parenchymal haemorrhagic
infarction (OR = 1.2, 31, and 17.6, respectively). Perinatal asphyxia is the only significant variable retained
for the prediction of severe CP versus mild or moderate CP.
Conclusion: The presented model based on perinatal characteristics and neonatal US-detected brain injuries
is a useful tool in identifying specific infants at risk for developing CP.
© 2010 Elsevier Ireland Ltd. All rights reserved.

1. Introduction
CP is one of the most common causes of motor disability in
childhood with a prevalence varying between 1.5 and 3 per 1000 live
births [1–4]. CP is defined as a disorder of movement and posture
causing activity limitations attributable to non-progressive distur-
bances occurring in the developing foetal or infant brain [5]. Cranial
ultrasonography (cUS) is an excellent tool to detect the most
frequently occurring brain abnormalities in preterm and full-term
neonates [6]. Brain lesions detected by cUS have proven to be
predictive for the development of CP [7–9]. However some children
with CP have normal US images. Other adverse perinatal elements
such as low gestational age, small for gestation, multiple gestation,
foeto-placental infection/inflammation, postnatal steroids, low
APGAR scores, duration of mechanical ventilation are all associated
with the development of CP [10–15]. However in the attempt to
⁎ Corresponding author. Rehabilitation Sciences and Physiotherapy Ghent, Campus
Heymans 2B3, De Pintelaan 185, BE-9000 Ghent, Belgium. Tel.: + 32 9 332 44 18;
fax: + 32 9 332 38 11.
E-mail address: eveline.himpens@ugent.be (E. Himpens).
unravel risk factors for the development of CP most studies rely on
univariate analyses or small numbers of infants followed-up. The most
recent study of the EPIPAGE study group (large population based
study on infants with a gestational age b 32 weeks) assessed the
independent role of US-detected lesions combined with several other
neonatal and obstetric factors as potential predictors of the develop-
ment of CP [16]. They concluded that cerebral lesions on ultrasound
are the most important predictors of CP in very preterm infants. In
addition, they found male sex and preterm premature rupture of
membranes/preterm labour are also independent predictors of CP in
very preterm infants. Absent in literature however is an individual
prediction model based on independent predictors of CP. Nonetheless,
having disposition of an individual prediction on the development of
CP would help neonatologists in the early decision-making process
and parental counselling, as well as prediction of the CP character-
istics would enable therapists to define more precisely short and/or
long term intervention strategies.
This study aims to pinpoint the predictive value of multiple
independent risk factors for CP including clinical perinatal variables
and neonatal cerebral ultrasound findings in a large consecutive
cohort of high-risk infants. More ambitiously, this study aims to create

0378-3782/$ – see front matter © 2010 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.earlhumdev.2010.05.019
414 E. Himpens et al. / Early Human Development 86 (2010) 413–417
a predictive model for the assessment of the individual risk of
developing cerebral palsy in a high-risk infant.
2. Methods
During an 11-year period (1995–2005) 1099 children were
assessed in the regional ‘Centre for Developmental Disorders Ghent’.
These children were referred from the NICU of the University Hospital
Ghent comprising all children with a GA less than 30 weeks and at risk
children with a GA ≥ 30 weeks with brain lesion and/or typical NICU-
related short and long term complications in serious degree (surgical
necrotising enterocolitis, chronic lung disease at 36 weeks PMA,
retinopathy of the prematurity grade III or more, meningitis, and
specific congenital infections). Assessment at the regional centre
consisted of a detailed physical, neurological and neurodevelopmen-
tal examination based on the quality of movement. Over the 11-year
period mainly one paediatric neurologist and two physiotherapists
were involved. The first and second assessment took place at the
corrected age of 4–6 and 12 months, respectively. In case of abnormal
or suspicious neurological findings or variant/deviant motor devel-
opment a third assessment took place at the corrected age of 2 years
or more (Table 1).
Medical records of each child were reviewed and the following risk
factors were recorded: gestational age (GA), small for gestational age
(SGA; BW b 10th percentile for GA [17]), gender, multiple gestation
(MG), perinatal asphyxia (PA) [18], duration of mechanical ventila-
tion (MV; b2 days, 2–7 days and N7 days) and cerebral US abnormal-
ities detected during the NICU stay. In NICU, serial cranial ultrasound
(US) were performed at least on days 1, 3, 7, 14 and 90 (or at the time
of discharge). During the 11-year period of data collection, a team of 7
neonatologists with extensive experience interpreted the US images.
Their conclusions were categorized by the first author. In case of
doubt the second author was consulted. Brain images were identified
as normal, white matter disease (WMD), intraventricular haemor-
rhage (IVH), cerebral infarction (CI), deep grey matter injury (DGM)
and parasagittal damage (PD). WMD was subdivided into transient
echodensities persisting less than 7 days, periventricular echodensi-
ties present for more than 7 days (PVL grade I), small localized cysts
(PVL grade II), and extensive periventricular or subcortical cystic
lesions (PVL grade ≥ III) [19]. Intraventricular haemorrhage com-
prised subependymal haemorrhage (IVH grade I), intraventricular
haemorrhage without or with ventricular enlargement (IVH grades II
or III, respectively), and parenchymal haemorrhagic infarction (IVH
grade IV) [20]. Cerebral infarction referred to necrosis confined to the
territory of one or more large cerebral arterial blood vessels. Deep
grey matter injury involved the thalamus and/or basal ganglia and
parasagittal damage referred to a lesion of the cerebral cortex and
Table 1
Follow-up data of the assessed cohort of 984 children.
CP No CP
Number of children 162 (16.5) 822
GA groups (weeks)
23–27 26 (16.6) 131
28–31 56 (14.3) 336
32–36 36 (16) 189
≥37 44 (21) 166
Timing of assessment
4–6 months 155 (95.6) 761 (92.6)
12 months 154 (95.1) 659 (80.2)
≥2 years 126 (77.7) 385 (46.8)
Median age first diagnosis CPa 12 – –
Diagnosis CP at 12 monthsb 73.3 – –
Data are numbers (%) unless mentioned otherwise; GA = gestational age.
a
Expressed in months of corrected age.
b
Expressed as percentage.
subcortical white matter with a characteristic parasagittal distribution
[21].
Subtypes and distribution of CP are defined according the SCPE
study group (Surveillance of Cerebral Palsy in Europe). Children were
classified as spastic or non-spastic CP (comprising dyskinesia and
ataxia). According the distributional aspect they were further
categorized into bilateral (diplegia and quadriplegia) or unilateral
spastic CP. Severity of CP is graded as mild + moderate (walking) or
severe (not walking). A child with a mild form of CP performs gross
motor skills independently but speed, balance and coordination are
reduced. His gait is rather clumsy than disabled. CP is defined as
moderate if the child reaches independent walking with or without
walking aids. His gait is obviously impaired. When the child has no
basic antigravity postural control and is unable to walk CP is classified
as severe [22,23].
Of all children referred from NICU 85% were assessed at the
regional ‘Centre for Developmental Disorders. Fifty infants with major
congenital malformations and/or karyotype anomaly and five chil-
dren with generalized hypotonia were excluded. Five children with
sub- or epidural bleeding and one child with sinovenous thrombosis
were excluded from the statistical analysis due to small numbers and
incompatibility with other categories. Another 54 children were
omitted due to missing data for at least one risk factor, leaving 984
children eligible for risk analysis (Table 1).
All statistical analyses were performed using SPSS 15 for Windows
(SPSS Inc.) and R Version 2.9.1 (package ROCR [24]). Univariate and
multivariate logistic regressions were performed to determine risk
factors for CP and its characteristics. Odds ratios (OR) were calculated
with 95% confidence intervals (CI). A model for the individual
prediction of CP was built using a stepwise backwards likelihood
ratio method. Variables were retained if significantly associated with
CP at the 5% significance level, but GA, gender and multiple gestation
were included in the model regardless of their significance because
they are generally accepted to be of influence. At different levels of
sensitivity, cut-off values for the predicted probability of CP, the
negative predictive value, the false negative and false positive rates
were determined. Sensitivity is the proportion of children with CP
who were identified as at risk for developing CP. The negative
predictive value is the proportion of patients with negative test results
(i.e., predicted probability of CP below a chosen cut-off) who are
correctly diagnosed. The false negative rate is the proportion of
patients with CP for whom the predicted probability of CP was below
a cut-off. The false positive rate is the proportion of patients without
CP for whom the predicted probability of CP exceeded the cut-off. The
performance of the prediction model was measured by the area under
the receiver operating characteristic (ROC) curve (AUC). Nonpara-
metric tests and 95% confidence intervals were used to compare AUC's
between models using the roc macro in SAS [25].
This study has been approved by the ethical committee of the
University Hospital Ghent.
3. Results
All
984 The perinatal characteristics of the 984 eligible children in relation
to CP are presented in Table 2. In this total cohort of high-risk infants
157 162 (16.5%) developed CP at follow-up.
392
225
210 3.1. Predicting CP
916 (93.1) Significant risk factors for CP identified by univariate as well as
813 (82.6)
multivariate analysis are perinatal asphyxia (PA), mechanical venti-
511 (51.9)
lation N 7 days (MV N 7 days), white matter disease except for tran-
sient echodensities b 7 days, IVH grades III and IV, cerebral infarction
(CI) and deep grey matter lesions (DGM). Of those, the US-detected
brain injuries are the major predictors. Within the groups of WMD
and intraventricular haemorrhage grading of brain involvement is
 E. Himpens et al. / Early Human Development 86 (2010) 413–417 415

Table 2
Univariate and multivariate regression model for the predictability of CP based on perinatal characteristics and neonatal neuroimaging data derived from a consecutive cohort of 984
high-risk infants.

CP No CP Univariate analysis Multivariate analysis

n (%) n (%)
OR (95% CI) P OR (95% CI) P

Gestational age (weeks) (GA) 1.0 (1–1.1) 0.05 1.1 (0.9–1.1) 0.05
Small for Gestational Age (SGA) 25 (15) 140 (85) 0.8 (0.6–1.4) 0.59 – –
Multiple Gestation (MG) 48 (17) 230 (83) 1.1 (0.7–1.6) 0.71 1.3 (0.8–2.1) 0.22
Gender (male) 95 (17) 450 (83) 1.2 (0.8–1.6) 0.41 1.2 (0.8–1.8) 0.44
Perinatal Asphyxia (PA) 32 (28) 81 (72) 2.2 (1.4–3.5) b0.001 2.4 (1.3–4.6) 0.008
Mechanical Ventilation (MV)
2–7 days 43 (16) 222 (84) 1.2 (0.8–1.8) 0.39 1.5 (0.9–2.5) 0.12
N7 days 47 (24) 150 (76) 1.9 (1.3–2.9) 0.002 3.1 (1.8–5.5) b0.001
White matter disease 80 (28) 203 (72)
Transient echodensities b7 days 7 (7) 90 (93) 0.6 (0.3–1.3) 0.19 1.1 (0.5–2.6) 0.86
PVL grade I 42 (31) 93 (69) 3.6 (2.2–5.2) b0.001 7.4 (4.4–12.4) b0.001
PVL grade II 8 (42) 11 (58) 5.4 (2.1–13.8) b0.001 10.9 (3.9–30.5) b0.001
PVL grade ≥ III 17 (85) 3 (15) 42.1 (12.1–146.8) b0.001 73.1 (19.8–270) b0.001
Haemorrhage 32 (20) 79 (80)
IVH grade I 6 (13) 39 (87) 0.9 (0.4–2.1) 0.76 1.4 (0.5–3.7) 0.46
IVH grade II 6 (16) 31 (84) 1.1 (0.4–2.7) 0.84 1.2 (0.4–3.3) 0.76
IVH grade III 10 (63) 6 (37) 9.4 (3.4–26.4) b0.001 24.2 (7.7–75.9) b.001
IVH grade IV 10 (77) 3 (23) 18.8 (5.1–69.5) b0.001 58 (14.8–226.3) b0.001
Cerebral infarction (CI) 14 (44) 18 (56) 4.2 (2.0–8.6) b0.001 9.2 (4–21.4) b0.001
Deep grey matter lesion (DGM) 18 (40) 27 (60) 3.7 (2–6.8) b0.001 4.1 (1.8–9.2) b0.001
Parasagittal damage (PD) 6 (26) 17 (74) 1.8 (0.7–4.7) 0.22 – –

P b 0.05; – = excluded from the model by the backwards stepwise method.

directly related to the likelihood of developing CP (Table 2). Under the (false negative) whereas 557 did not develop CP although they were
best fitting model, the risk of CP for an individual infant in a high-risk identified as such (false positive). The overall measure of performance
NICU population can be calculated as: of the presented prediction model, expressed by the area under the
ROC curve (AUC), is 83.3% (95% CI: 80–87%). A model based on the
egðCPÞ neuroimaging data alone results in an AUC of 79.5% (95% CI: 75–84%).
∏ðCPÞ =
1 + egðCPÞ The difference between the AUC's is significantly in favour of the
presented model based on US images and additional perinatal
gðCPÞ = −5:33 + 0 : 06 GA + 0:16 male + 0 : 29 MG + 1 : 15 ðMV N 7 daysÞ characteristics (−0.0378 (95% CI: − 0.0630 to − 0.0126); p = 0.003).
+ 0 : 89 PA + 1 : 41 DGM + 2 : 22 CI + 2 : 08 PVLI + 2 : 39 PVLII
+ 4 : 29 ðPVL≥IIIÞ + 3 : 19 IVHIII + 4 : 06 IVHIV
3.2. Predicting characteristics of CP
In case the risk factor is applicable on a specific infant, ‘1’ should be
The majority of CP children presents spastic bilateral CP. However
entered in the model. Otherwise the risk factor should be left out. GA
multivariate analysis revealed that with increasing gestational age or
should be entered in weeks. For example, for a male infant, born
in the presence of perinatal asphyxia and/or deep grey matter lesion,
singleton at NICU, with a gestational age of 28 weeks, IVH grade III but
it is more likely to develop the non-spastic form of CP (Table 4).
no other positive risk factors, we find
Concerning the distributional aspect of CP, multivariate analysis
g ðCPÞ = −5:33 + 0:06T28 + 0:16 + 3:19 = −0:3 revealed that with increasing gestational age and in the presence of
cerebral infarction and/or IVH grade IV it is more likely to develop
e−0:3 unilateral than bilateral spastic CP. Cerebral infarction is the leading
∏ðCPÞ = = 0:43 predictor of unilateral versus bilateral spastic CP [OR = 31 (95% CI: 4–
1 + e−0:3
242.6)] (Table 4).
indicating a 43% probability of developing CP. Cut-off values of the
prediction were determined corresponding to 95, 90 and 85% Table 4
sensitivities (Table 3). A sensitivity of 95% corresponded to a 4.5% Multivariate regression analysis on the predictability of subtype and distribution of CP.
cut-off meaning that 95% of all children with CP had a predicted
Subtype of CPa Distribution of CPb
probability for developing CP of at least 4.5%. Nine infants who
OR (95% CI) P OR (95% CI) P
developed CP were not identified as being at risk for developing CP
Gestational age (weeks) 1.1 (1–1.2) 0.03 1.2 (1–1.4) 0.03
Small for gestational age – – – –
Table 3 Multiple gestation – – – –
Cut-off values for the prediction of CP in relation to sensitivity, negative predictive Gender – – – –
value, false negative and false positive rates of the total cohort of 984 infants. Perinatal asphyxia 3.6 (1.2–10.9) 0.02 – –
Mechanical ventilation – – – –
Sensitivity Cut-off ∏(CP) N b cut-off NPV FN FP
White matter disease – – – –
(%) (%) N (%) (%) N (%) N (%)
Haemorrhage — IVH grade IV – – 17.1 (3.0–96.6) b.01
95 4.5 272 (28) 96.7 9 (5) 557 (68) Cerebral infarction – – 31 (4–242.6) b.01
90 6.4 456 (46) 96.3 17 (10) 381 (46) Deep grey matter lesion 7.5 (1.4–40.9) 0.02 – –
85 8.8 544 (55) 95.6 24 (15) 300 (37) Parasagittal damage – – – –
75 16.5 688 (71) 94.2 40 (25) 172 (21)
– = Excluded by the backward stepwise logistic regression method.
a
∏(CP) = probability for developing CP; NPV = negative predictive value; FN = false Reference category: spastic CP.
b
negative; FP = false positive. Reference category: bilateral spastic CP.
416 E. Himpens et al. / Early Human Development 86 (2010) 413–417
Concerning the severity of CP, stepwise logistic regression retained
perinatal asphyxia as the only significant predictor of severe CP versus
mild and moderate CP [OR = 5.2 (95% CI: 2.2–12.4)].
4. Discussion
This study focuses on a number of known risk factors to create a
model that computes the probability of development of CP in an
individual NICU-admitted infant. Gestational age, gender and multiple
gestation are not recognized as independent predictors in this study
although they are generally accepted as influencing factors and often
matched for in case–control studies. According to the literature, it is
more likely to develop CP being male [16,26] and/or being a part of a
multiple pregnancy especially in case of a dying co-twin [15,26]. In a
geographically based population, cerebral palsy rates increase with
decreasing gestational age. However, when infants are viewed as
potential candidates for cerebral palsy, the cerebral palsy rates seem
to increase with increasing gestational age [27]. This is in accordance
with the slightly increasing rate of CP in the more mature infants of
our high-risk cohort. Also small for gestational age is in this study not
recognized as an independent predictor of CP. Literature about this
risk factor is somewhat diverse. Poor intrauterine growth has been
shown to increase the risk of CP, particularly in moderately preterm
and term infants [10,11,28,29]. In extremely preterm infants, foetal
growth failure might reduce the risk of CP [11].
Perinatal asphyxia, mechanical ventilation for more than 7 days and
various types of brain US imaging anomalies are found to be independent
risk factors for the development of CP. Birth asphyxia is a rather
uncommon cause of CP [30]. However mild hypoxia in the presence of
other risk factors may amplify the risk of CP development [31,32]. The
odds ratio (OR) for perinatal asphyxia in the present study is 2.4 (95% CI:
1.3–4.6), whereas in the study of Greenwood et al. ORs for hypoxia are
much higher, respectively 12.2 (95% CI: 1.2–119) for infants with
GA≤32 weeks and 146.0 (95% CI: 7.4–3651) for term infants [11].
Mechanical ventilation (MV) has also been reported as an independent
risk factor for CP in the study of Drougia et al. [10] and in the study of
Laptook et al. [33] concerning extremely low birth weight infants. Several
reasons validate the association of MV and CP. MV produces positive
intrathoracic pressure that can impede venous return and thus decreases
cardiac preload. Subsequently, cardiac output can be compromised and
cerebral blood flow fluctuates [34]. Above all, brain injuries, especially
WMD, have been recognized in several studies as a significant risk factor
for CP [16,35–37]. In this study, thanks to the mixed population more
types of brain injury and their grading of involvement have been
recognized as major independent risk factors.
In this study, in accordance with the EPIPAGE study group, the
abnormal US images have the strongest predictive values. A model
based on US images alone however is significantly less accurate than
the prediction model on the basis of perinatal characteristics and US
data as demonstrated in this study. Other risk factors, reported in
literature or yet to be discovered, can and probably will play a role in
the development of CP. It is however inherent to each prediction
model and to the retrospective nature of this study in particular that
not all risk factors can be incorporated. The similar study of the
EPIPAGE study group also incorporated obstetric/maternal elements
in their multivariate analysis. However their contribution in the
prediction of CP was not significant.
In this study we concentrated strictly on cranial ultrasound images
to predict the development of CP since those data are very early in the
child's life at our disposal. However in clinical practice MR imaging is
often used to confirm or further specify US conclusions at some stage.
It would be interesting to compare the presented model with a model
based on additional MR imaging. In our study population, MRI has
been performed at some stage in only 15% of the children mostly term
infants and infants with obvious pathology born in the second half of
the study period. Due to this selection bias and the small numbers
statistical analysis on our MRI data would be inaccurate.’
In our study all the very preterm but only the more ill late preterm
and term infants were included. This selection bias does not distort
the reported associations between CP and its risk factors, because
odds ratios have the property to remain valid under outcome
selection. However, the intercept (−5.33) in the prediction model
is influenced by this selection and therefore less transportable to
patient populations that are otherwise selected. The prediction of the
development of CP for the individual NICU-admitted infant on the
basis of this model can provide a firmer basis for therapeutic decision-
making in neonates and help in parental counselling. The cut-off
points at different levels of sensitivity can be used depending on the
purpose. When parents are informed about the possibility of CP in
their infant from a predicted probability of developing CP exceeding
4.5% onwards, they might be needlessly worried since the false
positive rate at this cut-off point is high. On the other hand, at any
predicted probability under this cut-off value, parents can be
reassured since there is 96.7% certainty that such infant is not at
risk of CP. In terms of setting up an adequate and cost-efficient follow-
up program it should be considered to introduce a less sensitive cut-
off value with still a very high certainty not to develop CP. A slightly
lower cut-off value results in substantially more (almost 50%) infants
identified as ‘not at risk for CP’. Perhaps those children should not be
integrated in a follow-up program immediately but be assessed at the
ages of 3 and 6 when other developmental disorders as DCD, ADHD
and learning problems are taken with their appearance. A more
diversified follow-up program could offer the opportunity to screen
infants at risk at different stages of the development.
No previous studies performed risk analyses on perinatal data and
neonatal neuroimaging results in order to predict the characteristics of CP.
Classifying children with CP according their subtype, distribution and
severity yielded relatively small numbers limiting subgroup analysis.
Therefore, the results of this study need to be interpreted cautiously. Some
odds ratios of risk factors are accompanied with rather large confidence
intervals. However the presented results are in accordance with literature.
Asphyxia in term infants is related with deep grey matter injury which can
result in non-spastic CP [38,39]. Cerebral infarctions, as well as
haemorrhage grade IV are often unilateral brain lesions which can result
in unilateral neuromotor involvement [40].
In conclusion, the present study provides an innovative risk model
for the prediction of the development of CP based on perinatal
variables and neonatal neuroimaging data. Higher gestational age,
perinatal asphyxia and deep grey matter lesion are independent
predictors for the prediction of non-spastic CP relative to its spastic
counterpart. Independent risk factors for the prediction of unilateral
spastic CP are higher gestational age, cerebral infarction and
parenchymal haemorrhagic infarction (IVH grade IV). Perinatal
asphyxia is the only significant variable retained for the prediction
of severe CP relative to CP graded mild or moderate.
References
[1] Blair E, Stanley FJ. Issues in the classification and epidemiology of cerebral palsy.
Ment Retard Dev Disabil Res Rev 1997;3(2):184–93.
[2] Blair E, Watson L. Epidemiology of cerebral palsy. Semin Fetal Neonatal Med
2006;11(2):117–25.
[3] Nelson KB. Can we prevent cerebral palsy? N Engl J Med 2003;349(18):1765–9.
[4] Odding E, Roebroeck ME, Stam HJ. The epidemiology of cerebral palsy: incidence,
impairments and risk factors. Disabil Rehabil 2006;28(4):183–91.
[5] Surveillance of Cerebral Palsy in Europe (SCPE). Surveillance of cerebral palsy in
Europe: a collaboration of cerebral palsy surveys and registers. Dev Med Child
Neurol 2000 Dec;42(12):816–24.
[6] van Wezel-Meijler G, Steggerda SJ, Leijser LM. Cranial ultrasonography in
neonates: role and limitations. Semin Perinatol 2010 Feb;34(1):28–38.
[7] Ancel PY, Livinec F, Larroque B, Marret S, Arnaud C, Pierrat V, et al. Cerebral palsy
among very preterm children in relation to gestational age and neonatal
ultrasound abnormalities: the EPIPAGE cohort study. Pediatrics 2006 Mar;117
(3):828–35.
 E. Himpens et al. / Early Human Development 86 (2010) 413–417
[8] Himpens E, Oostra A, Franki I, Van Maele G, Vanhaesebrouck P, Van den Broeck C.
Predictability of cerebral palsy and its characteristics through neonatal cranial
ultrasound in a high-risk NICU population. Eur J Pediatr 2010.
[9] Kuban KCK, Allred EN, O'Shea TM, Paneth N, Pagano M, Dammann O, et al. Cranial
ultrasound lesions in the NICU predict cerebral palsy at age 2 years in children
born at extremely low gestational age. J Child Neurol 2009;24(1):63–72.
[10] Drougia A, Giapros V, KralliS N, TheochariS P, Nikaki A, Tzoufi M, et al. Incidence
and risk factors for cerebral palsy in infants with perinatal problems: a 15-year
review. Early Hum Dev 2007;83(8):541–7.
[11] Greenwood C, Yudkin P, Sellers S, Impey L, Doyle P. Why is there a modifying effect
of gestational age on risk factors for cerebral palsy? Arch Dis Child Fetal Neonatal
Ed 2005;90(2):141–6.
[12] Hagberg B, Hagberg G, Beckung E, Uvebrant P. Changing panorama of cerebral
palsy in Sweden. VIII. Prevalence and origin in the birth year period 1991–94. Acta
Paediatr 2001;90(3):271–7.
[13] Jacobsson B, Hagberg G, Hagberg B, Ladfors L, Niklasson A, Hagberg H. Cerebral
palsy in preterm infants: a population-based case–control study of antenatal and
intrapartal risk factors. Acta Paediatr 2002;91(8):946–51.
[14] Petrini JR, Dias T, McCormick MC, Massolo ML, Green NS, Escobar GJ. Increased risk
of adverse neurological development for late preterm infants. J Pediatr 2009;154
(2):169–76.
[15] Scher AI, Petterson B, Blair E, Ellenberg JH, Grether JK, Haan E, et al. The risk of
mortality or cerebral palsy in twins: a collaborative population-based study.
Pediatr Res 2002;52(5):671–81.
[16] Beaino G, Khoshnood B, Kaminski M, Pierrat V, Marret S, Matis J, et al. Predictors of
cerebral palsy in very preterm infants: the EPIPAGE prospective population-based
cohort study. Dev Med Child Neurol 2010 Feb 12.
[17] Devlieger H, Martens G, Bekaert A, Eeckels R, Vlietinck R. Perinatale activiteiten in
Vlaanderen; 1996.
[18] Sarnat HB, Sarnat MS. Neonatal encephalopathy following fetal distress — clinical
and electroencephalographic study. Arch Neurol 1976;33(10):696–705.
[19] De Vries LS, Eken P, Groenendaal F, van Haastert IC, Meiners LC. Correlation
between the degree of periventricular leukomalacia diagnosed using cranial
ultrasound and MRI later in infancy in children with cerebral palsy. Neuropedia-
trics 1993 Oct;24(5):263–8.
[20] Papile LA, Burstein J, Burstein R, Koffler H. Incidence and evolution of
subependymal and intraventricular hemorrhage: a study of infants with birth
weights less than 1, 500 gm. J Pediatr 1978 Apr;92(4):529–34.
[21] Volpe JJ. Hypoxic-ishemic encephalopathy. Neurology of the newborn. 5th ed.
Philadelphia: W.B. Saunders CO.; 2008.
[22] Wichers MJ, van der Schouw YT, Moons KGM, Stam HJ, van Nieuwenhuizen O.
Prevalence of cerebral palsy in The Netherlands (1977–1988). Eur J Epidemiol
2001;17(6):527–32.
[23] Balf CL, Ingram TTS. Problems in the classification of cerebral palsy in childhood. Br
Med J 1955;2(JUL16):163–6.
417
[24] Sing T, Sander O, Beerenwinkel N, Lengauer T. ROCR: visualizing classifier
performance in R. Bioinformatics 2005;21(20):3940–1.
[25] De Long ER, De Long DM, Clarke-Pearson DL. Comparing the areas under two or
more correlated receiver operating characteristic curves: a nonparametric
approach. Biometrics 1988;44:837–45.
[26] Bonellie SR, Currie D, Chalmers J. Comparison of risk factors for cerebral palsy in
twins and singletons. Dev Med Child Neurol 2005;47(9):587–91.
[27] Joseph KS, Allen AC, Lutfi S, Murphy-Kaulbeck L, Vincer MJ, Wood E. Does the risk
of cerebral palsy increase or decrease with increasing gestational age? BMC
Pregnancy Childbirth 2003 Dec 23;3(1):8.
[28] Gurbuz A, Karateke A, Yilmaz U, Kabaca C. The role of perinatal and intrapartum
risk factors in the etiology of cerebral palsy in term deliveries in a Turkish
population. J Matern-Fetal Neonatal Med 2006;19(3):147–55.
[29] Wu YW, Croen LA, Shah SJ, Newman TB, Najjar DV. Cerebral palsy in a term
population: risk factors and neuroimaging findings. Pediatrics 2006;118(2):
690–7.
[30] Jacobsson B, Hagberg G. Antenatal risk factors for cerebral palsy. Best Pract Res
Clin Obstet Gynaecol 2004;18(3):425–36.
[31] Keogh JM, Badawi N. The origins of cerebral palsy. Curr Opin Neurol 2006;19(2):
129–34.
[32] Kendall G, Peebles D. Acute fetal hypoxia: the modulating effect of infection. Early
Hum Dev 2005;81(1):27–34.
[33] Laptook AR, O'Shea TM, Shankaran S, Bhaskar B. Adverse neurodevelopmental
outcomes among extremely low birth weight infants with a normal head
ultrasound: prevalence and antecedents. Pediatrics 2005;115(3):673–80.
[34] Aly H. Mechanical ventilation and cerebral palsy. Pediatrics 2005;115(6):1765–7.
[35] Han TR, Bang MS, Lim JY, Yoon BH, Kim IW. Risk factors of cerebral palsy in
preterm infants. Am J Phys Med Rehabil 2002;81(4):297–303.
[36] Tran U, Gray PH, O'Callaghan MJ. Neonatal antecedents for cerebral palsy in
extremely preterm babies and interaction with maternal factors. Early Hum Dev
2005;81(6):555–61.
[37] Walstab JE, Bell RJ, Reddihough DS, Brennecke SP, Bessell CK, Beischer NA. Factors
identified during the neonatal period associated with risk of cerebral palsy. Aust
NZ J Obstet Gynaecol 2004;44(4):342–6.
[38] Pasternak JF, Gorey MT. The syndrome of acute near-total intrauterine asphyxia in
the term infant. Pediatr Neurol 1998;18(5):391–8.
[39] Van den Broeck C, Himpens E, Vanhaesebrouck P, Calders P, Oostra A. Influence of
gestational age on the type of brain injury and neuromotor outcome in high-risk
neonates. Eur J Pediatr 2008;167(9):1005–9.
[40] Roelants-van Rijn AM, Groenendaal F, Beek FJA, Eken P, van Haastert IC, De Vries
LS. Parenchymal brain injury in the preterm infant: comparison of cranial
ultrasound, MRI and neurodevelopmental outcome. Neuropediatrics 2001;32(2):
80–9.