This article is commented on by Hoon and Stashinko page 482 of this issue.
North American usage: mental retardation.
Dr Kate Himmelmann at Department of Paediatrics, Institute of Clinical Sciences, Queen Silvia Children’s Hospital, Sahlgrenska Academy at the University of Gothenburg, Gothenburg, Sweden. E-mail: firstname.lastname@example.org
Aim The aim of this population-based study was to describe function in cerebral palsy (CP) in relation to neuroimaging.
Method Motor function, accompanying impairments, and neuroimaging (86 by magnetic resonance imaging, 74 by computed tomography) were studied in 186 children born in western Sweden between 1999 and 2002 (96 males, 90 females; age range at data collection 4–8y).
Results Forty per cent of the children had unilateral spastic CP, 39% bilateral, 16% dyskinetic CP, and 5% ataxia. Fifty-one per cent were in level I of the Gross Motor Function Classification System (GMFCS), 14% in level II, 3% in level III, 11% in level IV, and 22% level V. Forty per cent of the children were in level I of the Manual Ability Classification System 19% were in II, 9% at III, 8% in IV, and 24% in level V. Seventy-six per cent of the children with white-matter lesions were in GMFCS levels I and II, whereas 67% with basal ganglia lesions were in levels IV and V. Learning disability* (45%), epilepsy (44%), and visual impairment (17%) were most common in children with brain maldevelopment, and cortical/subcortical or basal ganglia lesions. Speech was impaired in 49% of the children, absent in 30%, and 6% had a neuropsychiatric diagnosis. Compared with children born between 1991 and 1998, the numbers of those in GMFCS level I increased (p=0.007), as did those with epilepsy (p=0.015).
Interpretation Neuroimaging improves the understanding of the neuroanatomical basis for function in CP. Type and severity of motor impairment and accompanying impairments are related to the timing of lesions.
• This population-based study shows associations between neuroimaging and function in cerebral palsy.
• Basal ganglia and cortical/subcortical lesions carried the worst prognosis.
• Epilepsy and impaired cognition, behaviour, and communication constituted important additions to the motor impairment.
• A neuropsychiatric diagnosis was found in 6% and speech was impaired in half.
Children with cerebral palsy (CP) often have functional impairments and activity restrictions added to the motor impairment and, in the most recent definition by Bax et al.,1 problems relating to cognition, communication, behaviour, and sensation are recognized. Several studies have described how the accompanying impairments can be more disabling than the motor impairment.2–4 The occurrence of accompanying impairments such as epilepsy, learning disability, or impaired vision has been shown to correlate with the severity of the motor impairment.2–4 Communication may also be impaired5 and more subtle sensory and cognitive problems become apparent over time. There is also emerging evidence of neuropsychiatric and behavioural disorders.1,6,7 Therefore, screening for these conditions should be part of the assessment of a child with CP.8
In the past decade, improved neuroimaging has provided an opportunity to link the various CP types to specific aetiological periods and to the localization of the brain maldevelopment or lesions. Previous studies of the relation between motor function and neuroimaging have shown an association between focal periventricular white-matter lesions or focal cortical/subcortical lesions and less severe motor impairments, while children with a more diffuse encephalopathy, including basal ganglia lesions, were more likely to have severe motor impairments.9,10
Today, children with CP have other risk factors and a different survival rate compare to children from earlier birth cohorts. More children born extremely preterm survive. Despite this, half the children born before 27 weeks’ gestation still have severe neonatal morbidities, such as intraventricular haemorrhage and periventricular white-matter lesions, that will cause CP.11 Neonatal intensive care has improved, and so children with perinatal asphyxia, malformations, and various other severe conditions who would not have survived 10 years ago do so today.
The aim of this study was to describe function in children with CP in the light of neuroimaging findings and to depict trends over time in gross motor function, neuroimaging findings, and accompanying impairments.
The records of all 186 children with CP born from 1999 to 2002 who were living in a geographically defined area of western Sweden were reviewed.12 The age range of the children at data collection was 4 to 8 years. Nine children had a postnatal cause of the CP and were excluded from the analysis of neuroimaging. Neuroimaging had been performed in 160 of 177 children (74 with computed tomography [CT], 86 with magnetic resonance imaging [MRI]). Neuroimaging findings were classified into five categories: brain maldevelopment (Fig. S1, online only), periventricular white-matter lesions (Fig. S2, online only), cortical and deep grey-matter lesions, other lesions, and normal, according to Krägeloh-Mann.13 The group with cortical and deep grey-matter lesions was subdivided into cortical/subcortical (Fig. S3, online only) and basal ganglia (Fig. S4, online only) lesions. The subtypes of CP were classified according to the Surveillance of Cerebral Palsy in Europe14 into unilateral spastic, bilateral spastic, dyskinetic, or ataxic CP, by the dominating symptom. Diagnosis was made by the local paediatric neurologist. Any doubtful diagnosis was verified by one of the authors (KH). Preterm birth was defined as birth before 37 completed weeks of gestation. ‘Prenatal’ referred to the period of pregnancy until the onset of labour resulting in delivery. ‘Peri/neonatal’ was defined as the period from the onset of labour leading to delivery to the 28th day of life. Peri/neonatal adverse events were considered as cases of intracranial haemorrhage/stroke, cerebral infection, and hypoxic–ischaemic encephalopathy, not including children with prenatal cerebral anomaly or lesion.12 The Gross Motor Function Classification System (GMFCS),15 the Bimanual Fine Motor Function (BFMF) classification,16 and the Manual Ability Classification System (MACS)17 were applied to describe gross and fine motor functions respectively. The following accompanying neuroimpairments were recorded: learning disability, defined as mild in children with an IQ of 50 to 70 and severe if the IQ was <50 (IQ or developmental quotient measured by Wechsler18,19 or Griffith scales20 or estimated from clinical observation); epilepsy, defined as a diagnosis of epilepsy under treatment in the medical records; visual impairment, defined as an acuity of not more than 0.1 in the best eye with correction or the presence of functional blindness; hearing impairment, defined as the need for a hearing aid or the presence of deafness; a diagnosis of neuropsychiatric disorder in the form of attention-deficit–hyperactivity disorder (ADHD) and autism or autism spectrum disorders (ASDs; diagnosed according to DSM-IV21 by a paediatric neurologist or child psychiatrist).
Comparisons with children born between 1991 and 1998 for gross motor function, neuroimaging findings, and the occurrence of accompanying impairments were made to identify trends.2,15 The study was approved by the ethics committee at the Medical Faculty at the University of Gothenburg.
For contingency tables, the χ2 test for independence was used. A Kruskal–Wallis test was used to compare more than two groups. Fisher’s exact test was used in the case of small sample numbers, as indicated in the text. In the comparison between cohorts, the Cochran–Armitage χ2 test for trends was used.
Gross and fine motor function
Gross motor function, according to the GMFCS, was level I in 51%, level II in 14%, level III in 3%, level IV in 11%, and level V in 22% of the 186 children. The distribution varied between CP types (Fisher’s exact test, p<0.001; Fig. S5). Children with dyskinetic CP had the most severe motor impairment, with 87% in GMFCS levels IV and V (i.e. non-ambulant). Of the children with unilateral spastic CP and ataxia, most were independent walkers.
Fine motor function was classified according to the BFMF in 185 of the 186 children; 35% was in level I, 27% in level II, 11% in level III, 8% in level IV, and 18% in level V. The distribution of MACS levels, known in 155 children, was in level I in 40%, level II in 19%, level III in 9%, level IV in 8%, and level V in 24%. In 61% of 185 children, the gross (GMFCS) and fine motor (BFMF) functions were at corresponding levels.
Association of origin of CP, gestational age, and motor function
Children born at term with a prenatal origin of CP performed in GMFCS levels I or II in 38 of 52 (73%), whereas 23 of 43 (53%) of the children born at term with a suggested peri/neonatal origin of their CP were in GMFCS levels IV or V. Seventeen of the 26 (65%) children born at term with an Apgar score of <5 at 5 minutes and/or an arterial cord pH of <7.0 at birth, performed in GMFCS levels IV or V. Of the 62 children born preterm, the GMFCS levels were I or II in 39 (63%). Children in whom the origin of CP could not be classified (31/38, 82%) were in GMFCS levels I or II.
In the 115 children born at term (where there was a female predominance: 64 females, 51 males), there was a similar distribution of GMFCS levels between the sexes. In the children born preterm (where there was a male predominance: 37 males, 25 females), the males had a higher percentage in GMFCS levels IV or V than the females (χ2=5.7; p=0.017, GMFCS levels I–III and IV–V pooled; Fig. S6).
Neuroimaging findings, motor function, and CP type
The distribution of GMFCS differed between neuroimaging findings (Kruskal–Wallis χ2=15.8; p=0.001; Fig. 1), as did the distribution of fine motor function classification (Kruskal–Wallis χ2=15.6; p=0.001, BFMF levels pooled I–II and III–V; Fig. S7).
Twelve of 19 children with brain maldevelopment performed in GMFCS levels IV or V, whereas the remainder performed in levels I or II. Seven children had unilateral and seven bilateral spastic CP, whereas three had dyskinetic and two ataxic CP.
Of the 49 children with white matter lesions, which was the most common finding, 37 (76%) were in GMFCS levels I or II. Unilateral lesions were found in 27 of the 49 children. The 23 children with white-matter lesions born at term all performed in GMFCS levels I or II, compared with 14 of 26 of the children born preterm, whereas 11 were in levels IV or V.
Basal ganglia lesions
Twelve (67%) of the 18 children with basal ganglia lesions performed in levels IV or V and had a severe fine motor impairment. Most (13/18) had dyskinetic CP, whereas two had unilateral and three had bilateral spastic CP.
The 45 children with cortical/subcortical lesions, 14 of which were unilateral, performed either in GMFCS level I (20/45) or were severely motor impaired (17/45) in level V. The children in level I had unilateral spastic CP or bilateral spastic CP, apart from one child with ataxia, whereas those in level V had dyskinetic or bilateral spastic CP. Most (37/45) were born at term, 24 with a perinatal or neonatal origin of CP. Children with cortical/subcortical lesions were found at all levels of fine motor function.
Fourteen CT and 12 MRI findings were regarded as normal. Twenty-three of the 26 (88%) children with normal neuroimaging performed in GMFCS levels I or II. Children with normal CT or MRI findings most often had a mild fine motor impairment. Twelve children had bilateral spastic CP, eight had unilateral spastic CP, four had ataxia, and two dyskinetic CP.
In the 17 children who had not undergone neuroimaging, the distribution of GMFCS was 14 at I, two at level II, and one at level IV. Eight had unilateral spastic CP, seven bilateral spastic CP, and two had ataxia.
More than half the children had a cognitive function within the normal range (IQ>70; Tables I and Table SI, online only). Severe learning disability was most common in children with brain maldevelopment (present in 14/19), together with cortical/subcortical lesions (30/45; Fig. 2). Twelve per cent had a mild learning disability and one-third had a severe learning disability.
Table I. Accompanying impairments by type of cerebral palsy (CP) in 186 children born in western Sweden between 1999 and 2002
Unilateral spastic CP (n=74), %
Bilateral spastic CP (n=72), %
Dyskinetic CP (n=30), %
Ataxia, (n=10), %
Total (n=186), %
Each child may be counted in more than one category, and the percentages may add up to more than 100%.
Normal cognitive function
Mild learning disability
Severe learning disability
Severe visual impairment
Severe hearing impairment
Epilepsy was present in 44%. It was most common in children born after 32 weeks’ gestation, where almost half the children had epilepsy (Table SI). The percentage with learning disability and epilepsy varied by neuroimaging finding (Kruskal–Wallis χ2=19.9, p<0.001 and χ2=30.2, p<0.001 respectively) and was higher in children with brain maldevelopment or cortical/subcortical lesions on neuroimaging. Severe visual impairment was present in 32 children (17%), mostly in bilateral spastic and dyskinetic CP (Table I). In this group, 20 out of 31 (65%) had cortical/subcortical and/or basal ganglia lesions. Severe hearing impairment was recorded in six children (3%). Verbal communication was hampered in 91 (49%) of the children, 58 of whom had no speech. Of this latter group, 32 out of 58 (55%) had cortical/subcortical and/or basal ganglia lesions, whereas 10 (17%) had maldevelopment of the brain.
Nine of the 186 children had a diagnosis of autism or ASD (three with an IQ in the lower normal range, one with mild learning disability, and five with severe learning disability) and three had a diagnosis of ADHD, one of whom also had autistic features (one with an IQ in the lower normal range and two with mild learning disability). In the children diagnosed with autism, three had periventricular white-matter lesions, two had cortical/subcortical and basal ganglia lesions, imaging was normal in three, and there was malformation in one. One child with ADHD had a periventricular white-matter lesion, whereas the other two had normal findings. Five of the children with autism and all the children with ADHD were born at term. All but one of the 12 children with neuropsychiatric conditions were independent walkers.
Impairment load and neuroimaging
Cortical/subcortical lesions were the most prevalent neuroimaging finding in children with five or more accompanying impairments. Children with no or few accompanying impairments were more likely to have periventricular white-matter lesions or normal neuroimaging. Children with a mild motor impairment had fewer accompanying impairments, regardless of the neuroimaging findings. However, as many as 17 of 21 (81%) of those with cortical/subcortical lesions and in GMFCS levels I or II had accompanying impairments compared with 16 of 38 (42%) of the children with periventricular white-matter lesions or 12 out of 23 (52%) with normal neuroimaging findings. Children with unilateral cortical/subcortical lesions had more accompanying impairments than those with unilateral white-matter lesions.
Comparison with previous cohorts
Compared with children born between 1991 and 1998, the distribution of motor function expressed as GMFCS levels had changed, as well as the occurrence of epilepsy (Table II), whereas learning disability and severe visual impairment had not changed. For communication, 31% in the present cohort were non-verbal, which was about the same percentage as in the previous cohorts. Neuroimaging findings had a somewhat different distribution than in previous cohorts (Table II).
Table II. Trends in gross motor function, neuroimaging, and accompanying impairments in children with cerebral palsy born 1991 to 1998 and 1999 to 2002
Motor function, GMFCS level
1991–98 (n=367), %
1999–2002 (n=186), %
aCochran–Armitage χ2trend. bχ2 test for independence. GMFCS, Gross Motor Function Classification System.
The relationship of gross motor function to accompanying impairments has previously been described from the long-term study of CP in western Sweden.2,12,16,22 Some discernible patterns of functional impairments corresponding to neuroimaging findings could be shown for motor and non-motor functions in this study. There were also associations between CP type, aetiological period, neuroimaging, and function. In overall terms, there had been a shift in gross motor function towards the extremes of the GMFCS in the present cohort, with more children in levels I and V, compared with the distribution in children born between 1991 and 1998, and more like the findings from southern Sweden.23 Periventricular white-matter lesions were associated with a milder motor impairment and fewer accompanying impairments. The finding that periventricular white-matter lesions result in CP with a milder motor impairment in most cases agrees with the study by Robinson et al.10 In a recent report by Lasry et al.,24 children with white-matter lesions born preterm had a more severe motor impairment than those born at term. We found a similar difference, where all term-born children but only half of those born preterm were unaided walkers. It has been shown that axonal bypasses can be reorganized after periventricular lesions in the preterm brain.25 This may partly explain why children with periventricular white-matter lesions may be more mildly impaired. In contrast, children with basal ganglia lesions had a severe motor impairment, consistent with earlier findings.2 Children with cortical/subcortical lesions were either mildly motor impaired (GMFCS level I), or severely motor impaired (GMFCS level V), but with more accompanying impairments than children with other lesions, even among those with the mildest motor impairments. This can be only partly explained by the lesion being unilateral or by brain plasticity.25 Jacobson et al.26 recently reported a similar difference in visual field function, which was less affected in children with CP due to white-matter injury than that due to cortical/subcortical lesions. Children with normal neuroimaging in the present study were in GMFCS levels I or II in most cases, but more than half of them had accompanying impairments.
Exploring neuroimaging findings and relating them to functional aspects in population-based epidemiological studies further increases the opportunity to understand the neuroanatomical basis for function and the timing of lesions or interference with the developing brain. Previous studies have reported that most lesions can be assigned to the late second or third trimester, or to the perinatal period.10,22 Half the children born at term in this study had a suggested perinatal or neonatal origin of their CP and, among them, two-thirds had signs of asphyxia and performed in GMFCS levels IV or V.
The observed increase in epilepsy from 33 to 44% may be connected with the increase in cortical/subcortical lesions in the present cohort compared with earlier birth-year cohorts.
A new aspect that has not been explored in a systematic way in previous reports from the CP panorama of western Sweden is the occurrence of neuropsychiatric diagnoses. The amount of missing data is probably substantial. For this reason, the prevalences found in this study must be regarded as preliminary. Despite this, the finding of 6% with a diagnosis of a neuropsychiatric disorder, already at the age of 4 to 8 years, implies that this needs to be studied in more detail. In the same age group in the population, a prevalence of 6.2 per 1000 has been reported27 for ASDs alone. Behavioural and emotional symptoms were recently reported,28 as well as impaired attention and executive29 and social functioning.30 Another vital function, still not fully described in children with CP, is the ability to communicate. The occurrence of impaired communication in CP shown in this study was similar to that found by Andersen et al.5 in Norway. Classification scales for communication are underway. They will be a welcome contribution to the classification and measurement instruments currently available for describing function in CP in a more comprehensive way.
We conclude that more children were in GMFCS levels I and V than before. An increase in epilepsy to 44% was recorded, possibly because of an increase in cortical/subcortical lesions documented by neuroimaging. New facets of accompanying impairments were explored; they included neuropsychiatric disorders, found in 6%. Moreover, one-third of the children had no speech, and verbal communication was impaired in half. Periventricular white-matter lesions were associated with a mild motor impairment and fewer accompanying impairments, whereas brain maldevelopment and cortical/subcortical and basal ganglia lesions showed a reverse pattern. Children with cortical/subcortical lesions also had most accompanying impairments among the mildly motor impaired. More research is warranted to explore the many consequences of CP and its imaging correlates in greater detail. Neuroimaging improves the understanding of the neuroanatomical basis for function in CP, and should be performed in all cases. The type and severity of the motor impairment and the occurrence of accompanying impairments are related to the timing of lesions in the developing brain. The lesions are often associated with a clinical phenotype where specific needs may be anticipated and addressed. This is important for the planning of intervention in the child with CP.
We express our gratitude to research nurse Berit Askjung. This investigation was supported by grants from the Norrbacka-Eugenia Foundation, the AnnMari and Per Ahlqvist Foundation, the Torbjörn Jebner Memorial Foundation, the Västra Götaland Region, and the Folke Bernadotte Foundation. We also express our thanks to all our colleagues at the paediatric and habilitation departments in the study area.