Gross motor functional abilities in preterm-born children with cerebral palsy due to periventricular leukomalacia


* Correspondence to first author at Department of Neonatology, KE.04.123.1, Wilhelmina Children’s Hospital, PO Box 85090, 3508 AB Utrecht, the Netherlands.


To describe the impact of periventricular leukomalacia (PVL) on gross motor function, data on 59 children (37 males, 22 females) with a gestational age (GA) of 34 weeks or less with cerebral palsy (CP) due to PVL grade I (n=20), II (n=13), III (n=25), and IV (n=1) were studied; (mean GA 29wk 4d [SD 4wk 6d]; mean birthweight 1318g [SD 342]). Two independent raters used the Gross Motor Function Classification System (GMFCS) at four time points: T1, mean corrected age (CA) 9 months 15 days (SD 2mo 6d); T2, mean CA 16 months (SD 1mo 27d); T3, mean CA 24 months 27 days (SD 2mo 3d); and T4, median age 7 years 6 months (range 2y 2mo–16y 8mo). Interrater reliability and stability across time with respect to the total cohort were κ≥0.86 and ρ≥0.74 respectively. The association between PVL and gross motor outcome at T4 was strong (positive and negative predictive values 0.92 and 0.85 respectively). The proportion of children who remained in the same GMFCS level increased from 27% (T1–T4) to 53% (T2–T4) and 72% (T3–T4). PVL grade I to II, as diagnosed in the neonatal period, has a better functional mobility prognosis than PVL grade III–IV. These findings have implications for habilitation counselling and intervention strategies.

The main determinant for cerebral palsy (CP) in children born preterm is periventricular leukomalacia (PVL).1–6 PVL mainly occurs in infants born preterm with a birthweight of less than 1500g and between 24 and 34 weeks gestational age (GA).3,7,8 The pathogenesis of PVL is multifactorial.3,9

The prognosis of neuromotor development in children with PVL depends on the extent and localization of the white matter damage.2,4,8 PVL has a predilection for the transition of the corona radiata to the internal capsule where the corticospinal tracts are localized. Damage in this area may result in a typical clinical picture, often recognized and classified as ‘spastic diplegia’.2,4,9 As soon as the lesions extend laterally, quadriplegia may occur.2,8 Extensive lesions affecting the basal ganglia and/or the occipital area may also lead to cortical visual impairment 2,5,8,10,11 and/or epilepsy.11

Several studies have shown a relationship between PVL in general and subsequent development of CP.1–9,11 However, different PVL grades can result in different CP subtypes and outcomes. This phenomenon was reported earlier by our group1 and by Serdaroglu et al.11 Previous studies have reported other outcome classification systems at an early age,5,12 or have not focused exclusively on infants born preterm.11

To the best of our knowledge, there has not been a longitudinal study describing the relationship between the different PVL grades in infants born preterm and subsequent gross motor functional abilities classified with the Gross Motor Function Classification System (GMFCS). We, therefore, performed a cohort study of children born preterm with CP, to determine the association between the severity of PVL and the course and stability of gross motor abilities measured with the GMFCS at four time points between early infancy and childhood/adolescence.



The children who were eligible for this hospital-based follow-up study were born between 1990 and 2004 and discharged from a level three neonatal intensive care unit in the Wilhelmina Children’s Hospital, University Medical Center Utrecht (UMCU), the Netherlands, which serves a population of approximately 2 million inhabitants. Children were included who met the following criteria: a GA of 34 weeks or less, PVL identified by sequential neonatal cranial ultrasound, and a definite CP syndrome defined at 2 years of age.13 Children with the following conditions were excluded: a diagnosed genetic syndrome, a neuromuscular disorder, congenital anomalies, and less than 2 years of age at the time of the study. The parents were offered the opportunity to let their infants participate in a standardized follow-up programme from term age onwards until a rehabilitation programme was established. All parents were informed about data collection and consented to the study and ethical permission was obtained by the UMCU. Data were collected from January 1991 to May 2007.


Cranial ultrasound was performed soon after admission to the neonatal intensive care unit and thereafter at least once a week until discharge, and again during the first visit to the neonatal follow-up clinic around term age. Infants were scanned with an ATL UM-4 mechanical sector scanner with a multifrequency transducer (5, 7.5, and 10MHz crystals). For preference the 7.5MHz transducer was used for the best possible resolution. Cranial ultrasound abnormality was classified after its full evolution, which was either at discharge or at term age.1 Because there is no generally accepted classification system for PVL, we used the classification described by de Vries et al.8: PVL grade I, periventricular areas of increased echogenicity present for 7 days or more; grade II, periventricular areas of increased echogenicity evolving into small localized fronto-parietal cysts; grade III, periventricular areas of increased echogenicity evolving into extensive periventricular cystic lesions involving the occipital and fronto-parietal white matter; and grade IV, areas of increased echogenicity in the deep white matter evolving into extensive subcortical cysts. PVL grade IV is usually seen in more mature infants born preterm and in term infants after perinatal asphyxia.8

All scans were interpreted by a neonatal neurologist (LSdeV). The diagnosis of PVL was confirmed by magnetic resonance imaging (MRI): 52/59 (88.1%) children had at least one MRI, whereas 7/59 (11.9%) children had no MRI. Of the 36 children who had only one MRI, 10/59 children (16.9%) had the scan in the neonatal period and 26/59 (44.1%) in childhood. In 16/59 (27.1%) children MRI was performed in the neonatal period as well as in childhood.

Gross motor function

To classify the gross motor function according to the descriptions of the GMFCS,3,14–22 clinical notes in medical records were used with respect to the gross motor abilities of the children. The GMFCS describes the major functional characteristics of children with CP. It is a five-level pattern-recognition system. Children that are classified in GMFCS Levels I and II have the potential to walk independently both indoors and outdoors, and in the community as well. In contrast, children classified in GMFCS Levels III to V are limited in their self-mobility. They walk with a mobility device and are potential wheelchair users.

Use of the GMFCS requires no formal training.20 There are four age bands: before the second birthday, and at 2 to 4, 4 to 6, and 6 to 12 years of age. The GMFCS discriminates between children with CP syndromes according to their age-specific gross motor activity and is based on self-initiated movements.23 The GMFCS for children >2 years of age is reliable. The predictive value between the ages of 2 and 12 years is relatively stable over time (r=0.79).14,23 However, the reported interrater reliability for infants with CP who are less than 2 years old is moderate (κ=0.55).20 All children in the present study were, therefore, classified independently by two experienced paediatric physical therapists (ICvH and MJCE) at four age ranges, of which two were under the age of 2 years. In the event of discrepancies, consensus was reached by consulting a third person (JWG, a physician in paediatric rehabilitation medicine).

Data analysis

Data were analyzed with SPSS software (version 13.0). Because there was only one participant with PVL grade IV, we had to exclude this case from data analysis with regard to GA, birthweight, stability, proportion, and distribution of GMFCS levels across time. To determine whether the mean GA and the mean birthweight of the PVL grade I to III subgroups were comparable, one-way analysis of variance between groups was performed. Interrater reliability with respect to GMFCS levels was calculated as κ and its 95% confidence intervals (CI) with quadratic weighting for each period before correlations were calculated. Kappa statistics were categorized as poor agreement when lower than 0.20, as fair between 0.21 and 0.40, as moderate agreement between 0.41 and 0.60, as good agreement between 0.61 and 0.80, and as very good agreement above 0.80.24 The stability of GMFCS levels across time of the total cohort and of the PVL subgroups was determined by Spearman’s rank-order correlations and was expressed as both numbers and percentages. For clinical purposes we examined the prognostic value of PVL for gross motor outcome in childhood/adolescence by calculating the sensitivity, specificity, positive and negative predictive values, and odds ratio with 95% CI. Data for children with PVL were, therefore, dichotomized into grades I and II (mild) and III and IV (severe) and GMFCS Levels I and II and Levels III to V. A p value of less than 0.05 was considered to be statistically significant.


Participant characteristics

In all, 59 children born preterm (37 males, 22 females) met the inclusion criteria; complete GMFCS data were available for 57 children. Twenty children were diagnosed with PVL grade I, 13 with PVL grade II, 25 with PVL grade III, and one child with PVL grade IV. Their GA ranged from 26 weeks to 33 weeks 3 days (mean 29wk 4d [SD 4wk 6d]) and birthweight from 730 to 2140g (mean 1318g [SD 342]). There was no statistically significant difference between the PVL grade I to III subgroups with respect to GA and birthweight (p=0.70 and p=0.91 respectively). All children of our study cohort received paediatric physical therapy at some point in their life, and 49/59 (83%) of them were subsequently treated at a rehabilitation centre.

Gross motor function classification

The four periods were as follows: T1, mean corrected age (CA) 9 months 15 days (SD 2mo 6d), range 5 months 8 days to 12 months 23 days; T2, mean CA 16 months (SD 1mo 27d), range 12 months 23 days to 19 months 27 days; T3, mean CA 24 months 27 days (SD 2mo 3d), range 20 to 31 months; and T4, median age 7 years 6 months, range 2 years 2 months to 16 years 8 months. (The wide age range at T4 was due to the fact that the study period was up to May 2007; for some children we had a long follow-up period and for others the follow-up period was shorter.)

Interrater agreement, correlation of GMFCS levels across time, and stability

For 3/236 (1.3%) possible classifications we had missing data at T2 for one male with PVL grade I and at T2 and T3 for one male with PVL grade III.

Interrater reliability with respect to the GMFCS levels of the total cohort was very good: 0.86 (95% CI 0.75–0.97) at T1, 0.89 at T2, 0.94 at T3 and 0.96 at T4. Because of a substantial proportion of zeros in the data entries, the 95% CI could not be calculated for T2 to T4. The two raters disagreed in 53/233 (23%) of all ratings: at T1 in 15/59 (25%) ratings, at T2 in 16/57 (28%), at T3 in 13/58 (22%), and at T4 in 9/59 (15%). The difference between the raters was one level in 52/53 disagreements and two levels in one case. Correlations ranged between ρ=0.74 and ρ=0.90 (p<0.001) from T1 onwards with respect to the total cohort (the only participant with PVL IV was excluded). Within the PVL subgroups, correlations ranged from ρ=0.17 (p=0.48) to 0.88 (p<0.001; Table I). Table I also presents the number and percentage of children who remained in the same GMFCS level (stability) between T1 and T4, between T2 and T4 and between T3 and T4. Proportions increased from T1 onwards, with respect to both the total cohort and the PVL subgroups. In this part of Table I, the male with PVL grade IV is included in the total cohort (n=59).

Table I.   Correlations of Gross Motor Function Classification System (GMFCS) levels in children born preterm with cerebral palsy due to periventricular leukomalacia (PVL), from T1 to T4, and stability
PVL gradesCorrelation (ρ) of GMFCS levelsStabilityarelative to T4
  1. aNumber and percentage of children who remained in the same GMFCS level from T1, T2, and T3 (mean corrected ages) versus T4 (median age 7y 6mo); bcase with PVL grade IV included. ns, not significant.

I to III (n=58)
 T1 (9.5mo)0.76 (p<0.001)0.74 (p<0.001)0.74 (p<0.001)16/59b27.1
 T2 (16mo) 0.90 (p<0.001)0.88 (p<0.001)30/57b52.6
 T3 (24.9mo)  0.89 (p<0.001)42/58b72.4
I (n=20)
 T1 (9.5mo)0.53 (p=0.019)0.17 (ns)0.32 (ns)2/2010
 T2 (16mo) 0.64 (p=0.003)0.64 (p=0.003)10/1952.6
 T3 (24.9mo)  0.36 (ns)14/2070
II (n=13)
 T1 (9.5mo)0.33 (ns)0.42 (ns)0.35 (ns)2/1315.4
 T2 (16mo) 0.83 (p<0.001)0.88 (p<0.001)6/1346.2
 T3 (24.9mo)  0.84 (p<0.001)9/1369.2
III (n=25)
 T1 (9.5mo)0.54 (p=0.006)0.56 (p=0.004)0.51 (p=0.009)11/2544
 T2 (16mo) 0.87 (p<0.001)0.81 (p<0.001)13/2454.2
 T3 (24.9mo)  0.84 (p<0.001)18/2475

GMFCS levels across time

For the total study group better gross motor function, as expressed by GMFCS levels, was observed in 42/59 (71.2%) children at T4, in comparison with T1: 6/7 (86%) children with an initial GMFCS Level II, 18/23 (78%) with Level III, 12/19 (63%) with Level IV, and 6/10 (60%) with Level V; 16/59 (27.1%) showed a stable level and one child (1.7%) had to be classified from Level I to Level II.

Figure 1 presents the changing panorama of the GMFCS levels (median and 5–95% range) of the three PVL subgroups (grades I–III) from early infancy (T1) through to childhood/adolescence (T4).

Figure 1.

Gross Motor Function Classification System (GMFCS) levels from T1 to T4 in children born preterm with cerebral palsy due to periventricular leukomalacia (PVL). Box and whisker plots of the GMFCS levels according to PVL group (grade I, n=20; grade II, n=13; grade III, n=25) at T1, mean corrected age (CA) 9 months 15 days; at T2, mean CA 16 months; at T3, mean CA 24 months 27 days; at T4, median 7 years 6 months. The solid box shows 25th to 75th centiles; whisker bars extend to lowest and highest observed levels within 5 to 95% range. The bold horizontal line shows the median.

GMFCS levels in relation to severity of PVL

PVL grade I

All children with PVL grade I and CP (n=20) were classified in GMFCS Levels I and II at T4, with one exception who was classified in GMFCS Level III at T4. In more detail, the results show that 2/20 (10%) children remained within the same GMFCS level at all time points, whereas 18/20 (90%) children were classified in levels describing better gross motor abilities at T2 or T3 (Fig. 1).

PVL grade II

With regard to children with PVL grade II (n=13), none were classified in GMFCS Level V at T4: 9/13 (69.2%) were classified in GMFCS levels I and II, and 4/13 (30.8%) in Levels III and IV. In more detail, the results show that 1/13 (7.7%) children remained within the same GMFCS level at all time points, 10/13 (76.9%) children were classified in levels describing better gross motor abilities at T2 or T4. Two children (15.4%) were classified from Level II (T3) to Level III (T4) and from Level III (T3) to Level IV (T4) respectively.

The chance that a child with PVL grade I or II would subsequently be classified in GMFCS Levels III to V at T4 was 15% (negative predictive value 0.85; Table II).

Table II.   Association between periventricular leukomalacia (PVL) grade and gross motor function at T4 in children born preterm with cerebral palsy
PVL gradesGMFCS levelsTotal
  1. GMFCS, Gross Motor Function Classification System. T4, median age 7y 6mo. Sensitivity, 0.83; specificity, 0.93; negative predictive value, 0.85; positive predictive value, 0.92; odds ratio for PVL grades III and IV, 67.2 (95% confidence interval 11.9–378.3).

III and IV24226
I and II52833

PVL grade III

At T4, 2/25 (8%) children with PVL grade III were classified in GMFCS Levels I and II and 23/25 (92%) in GMFCS Levels III–V. In more detail, the results show that 8/25 (32%) children remained within the same level at all time points, whereas 3/25 (12%) were classified in GMFCS Level IV at all time points except one. A change to GMFCS levels describing better gross motor abilities was observed in 14/25 (56%) children: 11 children shifted one level from T1 to T4 and three children shifted two levels. One female was initially classified in GMFCS Level III but from T2 onwards in Level II, and one male, who had PVL more localized in the frontal area of the brain, shifted from GMFCS Level II at T1 and T2 to Level I from T3 onwards.

PVL grade IV

The male with PVL grade IV was classified in GMFCS Level V from the beginning and throughout the study period.

The chance that a child with PVL grade III or IV was subsequently classified in GMFCS Levels III to V at T4 was 92% (positive predictive value 0.92), with an odds ratio of 67.2 (95% CI 11.9–378.3; Table II).

Distribution of PVL severity related to CP subtype

At 2 years of age, a minority of the study cohort was diagnosed with a spastic unilateral CP and the majority with a spastic bilateral CP with more involvement of the lower extremities than the upper extremities (Table III). It is noteworthy that, at T4, three children with a PVL grade I and one female with a PVL grade II no longer met the criteria of CP,13,25 although they still showed other (motor) problems.

Table III.   Distribution of periventricular leukomalacia (PVL) severity related to cerebral palsy (CP) subtypes diagnosed at 2 years of age in children born preterm
PVL gradesCP subtypeTotal
  1. SU, spastic unilateral; SBD, spastic bilateral diplegia; SBQ, spastic bilateral quadriplegia.

I and II33033
III and IV121426
Total (%)3 (5.1)42 (71.2)14 (23.7)59 (100)


With regard to the total cohort, 35.6% of the children were able to walk independently, 40.7% could walk with an assistive device, and 23.7% could not walk at all and were wheelchair users at T4 (Table IV).

Table IV.   Periventricular leukomalacia (PVL) severity related to ambulation of children born preterm with cerebral palsy
PVL grades1–2y2–4y4–6yAssistedNot walkingTotal
  1. The children between the ages of 1 and 6 years walked totally independently. Assisted, walking with device; Not walking, totally dependent on wheelchair (and adapted bicycle).

I and II11813133
III and IV11111326
Total (%)12 (20.3)8 (13.6)1 (1.7)24 (40.7)14 (23.7)59 (100)

Ambulation in relation to PVL grade

The chance that a child with PVL grade I or II would walk independently was 93% (specificity; Table II). Of the children with PVL grades I and II, 19/33 (57.6%) were able to walk unassisted before preschool age, 13/33 (39.4%) could walk with an assistive device and one child (3%) was not yet able to walk at the age of 3 years 8 months (Table IV).

The chance that a child with PVL grade III or IV would not be able to walk in childhood/adolescence was 83% (sensitivity; Table II). One male (4%) with a PVL grade III was able to walk unassisted at 19 months CA and one female (4%) at 5 years 6 months. Eleven out of 25 (44%) children could walk with a device. Three of them could walk indoors but were dependent on assistive mobility aids in the community or needed a wheelchair for longer distances. Twelve out of 25 (48%) children were totally dependent on a wheelchair, of whom two used an adapted bicycle in addition to their wheelchair. The male with PVL grade IV was totally dependent on a wheelchair (Table IV).


The gross motor abilities of children born preterm who developed CP as a result of PVL vary depending on the severity of the PVL. Most infants with PVL grade III and IV were never classified in GMFCS Levels II or I and, therefore, did not achieve the potential to walk independently. This is in contrast with children with PVL grade I, and to a lesser extent PVL grade II, in which most of the children were able to walk. These findings are supported by the studies of Rogers et al.2 and Serdaroglu et al.11

The present study emphasizes the importance not only of studying children with a specific type of brain lesion, but also of studying them in more detail by subdividing them into different grades of PVL, to obtain a better clinical picture than by considering them as one undifferentiated group.

The interrater reliability in the present study was very good and in agreement with previous studies,14,19 even with regard to infants younger than 2 years of age.20 This applies to the total cohort and the subgroups with PVL grades II and III in particular. The use of the GMFCS in children with PVL grade I yielded less consistent results before 2 years of age. Rating of their gross motor function can, therefore, best be determined by a consensus of at least two examiners.

Stability of gross motor function, as expressed by GMFCS levels, became more robust after the first year of age. However, the results suggest that it is possible to predict gross motor outcome in children with PVL before this age. Our findings are consistent with the observation that children’s gross motor function changes with age and experience13,14 when it comes to children with PVL grades I and II. Palisano et al.19 demonstrated that, across time, 72% of the children remained at the same GMFCS level. With regard to our study group this applies to T3. As with Palisano et al. we showed that children initially classified in the extreme levels (I and V) were least likely to be reclassified.19 Our findings are also comparable to those of Gorter et al.18 and Romeo et al.,26 showing that children with a spastic unilateral type of CP do have better gross motor abilities than children with a spastic bilateral type in which all extremities are involved. Children with a spastic bilateral type, with more involvement of the lower extremities than the upper extremities, showed the most diverse gross motor clinical picture.

By comparing GMFCS levels across time with the severity of PVL and type of CP in children born preterm, the present study adds to the construct- and criterion-related validity of the GMFCS.

Several explanations for interobserver differences can be identified: (1) the quantity and quality of information available in the clinical notes,23 more specifically a lack of functional descriptions in the clinical records; (2) overlap in descriptions between GMFCS levels without clear distinctions; (3) different weighting by the raters of various aspects within the descriptions of the GMFCS23; and (4) a lack of clear definitions such as ‘running’ and ‘jumping’. We noticed that, over time, some children had to be reclassified in a level describing more limited gross motor abilities, although there was no deterioration of their functional capacities. This applied in particular to the transition from age category 0 to 2 years to age category 2 to 4 years.

One of the limitations of this study concerns the subgroup of infants with PVL grade I. Only infants with PVL and subsequent development of CP were eligible for the study. However, it is known from several studies that only 4 to 10% of infants with a cranial ultrasound diagnosis of PVL grade I will go on to develop CP. It is also well known that this form of white matter injury is more reliably diagnosed with MRI than with cranial ultrasound.27–29 A further possible limitation of the study is that we did not take into account the existing comorbidities (visual and/or hearing problems, epilepsy, learning difficulties) and interventions (type of therapy and spasticity treatment) of the children as predictors of GMFCS level, although most of them were seen in the regional rehabilitation centres on a regular basis. Another limitation might be that 13 participants were older than 12 years of age and 11 of these could have been considered adolescents at T4. The GMFCS for the adolescent population has not yet been published. However, the predictive values in our study support the observations of McCormick et al.,23 who stated that ‘the GMFCS level observed around 12 years of age is highly predictive of adult motor function’. It is also true to say that we do not know whether the gross motor function of the nine participants who were rated in GMFCS Levels II to V and who were between 2 and 4 years of age at T4 will improve; neither do we yet know whether the gross motor function of some of the children in our cohort will deteriorate with age.

In infants born preterm with PVL it is important to know the severity of the brain damage. Children with PVL grades I and II have the potential to walk independently and thus may find it easier to participate in their community. Children with PVL grades III and IV are more dependent on environmental modifications for daily activities, on mobility devices, and on support from others. The choice for specific indoor and/or outdoor mobility devices is dependent on the personal preferences of the caregivers and the child, and on environmental circumstances.22,30


This study contributes to a better understanding of the impact of PVL on gross motor functional abilities in children born preterm by demonstrating that a ‘composite diagnosis’ (neuroimaging and GMFCS) is worthwhile to obtain a clear clinical picture. The present study also offers opportunities to inform parents of the early and future functional possibilities of their child, and to choose the most appropriate intervention strategy.


We are grateful to our secretaries Mrs I G Borger-Gerritse and Mrs L Brouwer-Prins for their support, and to Dr E J H Mulder for his help in creating Figure 1.