Cognitive outcome in childhood after unilateral perinatal brain injury

Authors


Correspondence to Professor Dr Linda S de Vries, Department of Neonatology, Wilhelmina Children's Hospital, UMC Utrecht, KE 04.123.1, PO Box 85090, 3508 AB Utrecht, the Netherlands. E-mail: l.s.devries@umcutrecht.nl

Abstract

Aim

The aim of the study was to assess cognitive outcome in children with periventricular haemorrhagic infarction (PVHI) or perinatal arterial ischaemic stroke (PAIS) and relate these findings to early developmental outcome and neonatal magnetic resonance imaging findings.

Method

A neuropsychological assessment was performed in 50 children (26 males, 24 females) with unilateral PVHI (n=21) or PAIS (n=29) at a median age of 11 years 9 months (range 6–20y). This included tests for intelligence, verbal memory, visual–motor integration, word comprehension, attention, reaction times, and executive function. The Griffiths Mental Development Scale was used for early developmental assessment at 24 months (range 18–32mo).

Results

In children with PVHI, both the early Griffiths scores (mean 87; 95% CI 83–92) and the Full-scale IQ (FSIQ) scores at school age (mean 86; 95% CI 78–94) were below the test mean of 100. In the PAIS group, early Griffiths scores were within the normal range (mean 98; 95% CI 93–104), but at school age FSIQ scores were below average (mean 87; 95% CI 80–94). In children with PVHI, FSIQ scores correlated with the level of maternal education and were lower after ventricular dilatation, whereas both involvement of the basal ganglia and thalami and development of postneonatal epilepsy were associated with lower cognitive outcome in children who had experienced PAIS.

Interpretation

Cognitive outcome after PVHI or PAIS is below average, but still within 1SD for most children. Prediction of cognitive outcome remains challenging, but some early predictors can be recognized.

Abbreviations
FSIQ

Full-scale IQ

PAIS

Perinatal arterial ischaemic stroke

PVHI

Periventricular haemorrhagic infarction

PHVD

Posthaemorrhagic ventricular dilatation

What this paper adds

  • Cognitive outcome after PVHI is below average but stable over time.
  • Posthaemorrhagic ventricular dilatation requiring intervention is associated with a worse cognitive performance.
  • Lower cognitive performance after PAIS is associated with involvement of the basal ganglia and thalami and development of postneonatal epilepsy.

Perinatal brain injury is commonly encountered in the neonatal intensive care unit in both preterm and term-born infants. Different types of injury can be observed and depend on the gestational age at which they arise.[1] Periventricular haemorrhagic infarction (PVHI) is a typical example of white matter injury which may occur during the early third trimester. Periventricular haemorrhagic infarction is considered to be the result of impaired drainage of the medullary veins in the periventricular white matter and tends to be preceded by an ipsilateral intraventricular haemorrhage or germinal matrix haemorrhage.[2] Towards the end of the third trimester, the pattern of injury changes to one of predominantly cortical and subcortical grey matter injury. Perinatal arterial ischaemic stroke (PAIS) is an example of such injury and is caused by a focal disruption of cerebral blood flow due to arterial embolization or thrombosis, resulting in focal ischaemia.[3]

Both PAIS and PVHI often result in adverse neurological sequelae in childhood, including motor and cognitive impairment.[3] While motor deficits can usually be predicted based on the size and site of the lesion,[4] early prediction of cognitive development remains a challenge. Studies reporting on cognitive function after PVHI or PAIS have reported different outcomes, ranging from normal performance to cognitive delay.[5-8] Most studies, furthermore, reported cognitive outcome before school age, while it has been suggested that cognitive deficits may not become evident until school age.[8]

The aim of the current study was to assess cognitive outcome at school age in a cohort of children with unilateral PVHI or PAIS. We compared the cognitive data with data of early developmental assessments. Furthermore, we tried to correlate characteristics of the lesion on neonatal magnetic resonance imaging (MRI) with cognitive outcome.

Method

Participants

Participants were selected from a database of children with unilateral PVHI and PAIS admitted to the neonatal intensive care unit of the Wilhelmina Children's Hospital, Utrecht between 1991 and 2005. Children were recruited for a study on brain plasticity after perinatal brain injury. Three additional children with PAIS were recruited for this study from the VU University Medical Centre and the results of their neuropsychological assessment were also included in this study. This resulted in a total of 50 children (26 males, 24 females) with PVHI (n=21) and PAIS (n=29).

Clinical data

Demographic and neurological characteristics were obtained from the children's charts and via a parental questionnaire. Maternal educational level was used as an indicator of socio-economic status and rated on a seven-point scale according to Verhage,[9] where higher scores reflect higher levels of education. Education of the child was categorized as either special or mainstream education. The diagnosis of unilateral spastic cerebral palsy (CP) was made according to the criteria of Bax et al.[10]

Neuroimaging data

Magnetic resonance imaging was performed at term-equivalent age in preterm children or during the first week after birth in term-born children. As a result of the inclusion of children born over a long period, the MRI protocol was not the same for all children. All MRI (acquired on a 1.5T MR system, Philips Medical Systems, Best, the Netherlands) did, however, include sagittal T1-weighted and axial T2-weighted imaging, as well as inversion-recovery T1-weighted imaging and often diffusion-weighted imaging. In two children, no neonatal MRI was carried out and PVHI and PAIS were diagnosed using cranial ultrasound. These two children were scanned beyond the neonatal period using a similar MRI protocol.

In children with PVHI, the site of the PVHI was noted. The presence of posthaemorrhagic ventricular dilatation (PHVD) requiring intervention (lumbar punctures or insertion of a ventricular reservoir or a ventriculoperitoneal shunt) was recorded from the infants' charts. After PAIS, the involved branch was noted. In these children, we additionally recorded whether the cortex or basal ganglia and thalami were involved.

Early developmental assessment

Children who were admitted to the neonatal intensive care unit of the Wilhelmina Children's Hospital (n=47) were seen at regular intervals in the neonatal follow-up clinic. In these children, an early outcome assessment was performed at 24 months (range 18-32mo) using the Griffiths Mental Developmental Scale.[11] Depending on the child's age, this test consists of five or six subscales (locomotor, personal–social, language, eye–hand coordination, performance, and practical reasoning), resulting in a mean score of 100 and an SD of 15. The developmental quotient was calculated using all subscales excluding the locomotor scale to avoid lower scores in children with motor deficits but with normal cognitive performance. The scores of preterm children were corrected for gestational age up to the age of 24 months.

Neuropsychological assessment

At school age, different neuropsychological functions were tested. First, intelligence was tested with either the Wechsler Intelligence Scale for Children[12] or the Wechsler Adult Intelligence Scale,[13] depending on the child's age. Scores derived included Verbal IQ, Performance IQ, and Full-scale IQ (FSIQ) with a mean score of 100 and an SD of 15.

In addition, we assessed verbal memory, visual–motor integration, word comprehension, attention, reaction times, and executive function in everyday life to investigate whether these neuropsychological functions were specifically impaired in this group of children.

Verbal memory was tested with the 15-word list,[14] which consisted of five learning trials with immediate recall of words (resulting in a total score) and a delayed recall after 25 minutes. Visual–motor integration was assessed using the Beery Developmental Test of Visual–Motor Integration,[15] with a developmental sequence of geometric forms to be copied with paper and a pencil. Word comprehension was tested with the Peabody Picture Vocabulary Test.[16] Attention and concentration in terms of speed and accuracy were assessed using the Bourdon–Vos Test[17] and scored on a three-point scale. Reaction times were examined by two tasks:[18] one visual task and one auditory task. Finally, executive function was assessed using the Behavioral Rating Inventory of Executive Function, which is a parental questionnaire regarding well-organized, purposeful, goal-directed, and problem-solving behaviour.[19] The Behavioral Rating Inventory of Executive Function comprises 86 items on two subscales: behavioural regulation and metacognition. A global executive composite was also calculated.

Statistical analyses

Test results were compared with the theoretical mean of the normative sample using a one-sample t-test or a Wilcoxon signed-rank test where appropriate. Differences between groups were evaluated using independent-samples t-tests and changes in intellectual ability of the groups over time were evaluated using paired t-tests. For individual data, a change in intellectual ability over time was considered significant if the observed difference was larger than two times the standard error of measurement (SEM), reflecting the test–retest reliability. As different tests were used at the two time points, the more stringent SEM of the Wechsler Intelligence Scale for Children was used (3.48 points).

Finally, univariate analyses were performed using a two-sample t-test or a Spearman's rank correlation coefficient, followed by forward linear regression if possible. All data were analyzed using the IBM SPSS Statistics version 20 (IBM Corp., Somers, NY, USA).

Informed parental consent was obtained and agreement was obtained from children aged 12 years and older. The institutional review board of the University Medical Centre Utrecht approved the study.

Results

Clinical characteristics

The clinical characteristics of the children with PVHI (n=21) and PAIS (n=29) are shown in Table 1. As expected, mean gestational age was lower in children with PVHI than in children with PAIS. Unilateral spastic CP was observed in 26 children. Thirty-eight children (76%) attended mainstream education. The handedness did not differ between both groups.

Table 1. Demographic and neurological information
VariablePeriventricular haemorrhagic infarction (n=21)Perinatal arterial ischaemic stroke (n=29) p
  1. NS, not significant.

Sex (male/female)14/712/17NS
Gestational age (range)306/7wks (26–41wks)38wks (295/7–425/7wks)<0.001
Handedness (left/right)(12/9)(12/17)NS
Localization (left/right)12/920/9NS
Unilateral spastic cerebral palsy (%)12 (57)14 (48)NS
Postneonatal epilepsy (%)2 (10)7 (24)NS
Education of the child (mainstream/special)15/623/6NS

Nine children developed epilepsy at a median age of 6.5 years (range 2.5–13y). Seizures were more often observed in children with PAIS (n=7) than in children with PVHI. Most children suffered simple partial seizures (n=7), which extended into secondary generalized seizures in two children. Of the remaining two children, one child suffered symptomatic multifocal seizures and one child's EEG showed electric status epilepticus in slow-wave sleep. One antiepileptic drug was used in seven children for seizure control, while two children required two antiepileptic drugs.

Neuroimaging data

Periventricular haemorrhagic infarctions were located in the frontal region of the brain in six children, in the frontoparietal region in 14 children and in the parietal region in one child. Most lesions were located in the left hemisphere (n=12). Nine children were treated for PHVD using lumbar punctures. Three eventually required a ventricular reservoir, while three others required a ventriculoperitoneal shunt.

Perinatal arterial ischaemic stroke was most frequently noted in the middle cerebral artery territory (n=26; in the main branch in eight children, in the anterior branch in two children, in the posterior branch in one child, in the lenticulostriate branch in seven children, in the cortical branch in seven children, and in the distal middle cerebral artery branch in one child). The three remaining children showed a PAIS in the posterior (n=2) and anterior (n=1) cerebral artery territory. The left hemisphere was more often affected (n=20). In the majority of individuals, the cortex was involved (n=23). Involvement of the basal ganglia and thalami was observed in 16 children. In eight children, additional injury was observed in the contralateral hemisphere, involving subtle white matter injury.

Early developmental assessment

An early developmental assessment was performed in 47 children, at a median age of 24 months (range 18–32mo). Lower scores were observed in children with PVHI (mean 87; 95% CI 83–92). Among children with PAIS, performance on the Griffiths Mental Developmental Scale did not differ from that of the normative sample (mean 98; 95% CI 93–104), and was significantly higher than in children with PVHI (mean difference 11; 95% CI 4–18). When subdividing the PAIS group into preterm and term children, no significant difference in developmental quotient was observed between these children.

Neuropsychological assessment

A neuropsychological assessment was performed at a median age of 11 years 9 months (range 6–20y). Compared with the normative sample, the PVHI group exhibited significantly lower performance on the FSIQ (mean 86; 95% CI 78–94) and Performance IQ (mean 80; 95% CI 73–87), but not on the Verbal IQ (mean 94; 95% CI 86–103) while the PAIS group showed below average performance on the FSIQ (mean 87; 95% CI 80–94), Performance IQ (mean 86; 95% CI 79–93), and Verbal IQ (mean 90; 95% CI 84–96).

Verbal memory (15-word list) and word comprehension did not differ from the normative sample for both groups. Handedness and location of the lesion did not influence word comprehension. Visual–motor integration was significantly lower in both the PVHI (mean 76; 95% CI 70–82) and the PAIS (mean 86; 95% CI 81–92) groups than in the normative sample, and was lower in the PVHI group than in the PAIS group (mean difference 11; 95% CI 3–19). Attention and concentration in terms of speed and accuracy were found to be poorer in children with PVHI. Of the children with PVHI, 67% had a poor speed performance (compared with 24% of the children with PAIS), and 48% had a poor performance on accuracy (compared with 24% of the children with PAIS). In both groups reaction times in response to visual stimuli were significantly shorter than in the normative sample (PVHI: mean 77; 95% CI 65–88; PAIS: mean 75; 95% CI 63–86). In contrast, no significant difference in reaction times in response to auditory stimuli was observed. Parent ratings of executive function (Behavioral Rating Inventory of Executive Function) were not significantly higher for both groups. Table 2 shows the test results for each domain.

Table 2. Test results
MeasureVariableTest meanPeriventricular haemorrhagic infarction (n=21)Perinatal arterial ischaemic stroke (n=29)
n Sample meanSD p n Sample meanSD p
  1. DQ, developmental quotient; WISC, Wechsler Intelligence Scale for Children; WAIS, Wechsler Adult Intelligence Scale; FSIQ, Full-scale IQ; VIQ, Verbal IQ; PIQ, Performance IQ; 15WT, 15-word list; Delayed (1), corrected for total score child; Delayed (2), corrected for sex and age; Beery, Beery Developmental Test of Visual-Motor Integration; PPVT, Peabody Picture Vocabulary Test; WCQ, Word Comprehension Quotient; RT, reaction time; BRIEF, Behavioral Rating Inventory of Executive Function; BRI, behavioural regulation index; MCI, metacognition index; GEC, global executive composite.

GriffithsDQ1002087.409.790.0002798.1513.180.472
WISC, WAISFSIQ1002186.1417.050.0012987.1717.880.001
VIQ1002194.0518.590.1582990.2116.320.003
PIQ1002179.8614.610.0002986.2418.910.001
15WTTotal5215.863.550.282274.302.850.211
Delayed (1)5215.103.130.890274.783.040.707
Delayed (2)5215.483.120.493274.262.890.195
Beery 1002175.7113.200.0002786.4114.530.000
PPVTWCQ10021103.7115.010.2702697.4616.180.431
RTVisual502176.8125.190.0002674.8828.520.000
Auditory502048.7527.170.8392447.8827.950.713
BRIEFBRI502054.8035.410.5522646.7335.790.645
MCI502045.8527.310.5052634.0028.880.009
GEC502048.7531.450.8612637.3832.870.062

Relation between early and late assessment

In the PVHI group, overall intellectual ability at school age was not significantly poorer than at the early developmental assessment (mean difference 2; Fig. 1). In the PAIS group, overall intellectual ability as measured by the Griffiths developmental quotient was significantly lower at school age than at the early developmental assessment (mean difference 11; 95% CI 5–17; Fig. 1).

Figure 1.

Developmental outcome (Griffiths developmental quotient, white bars) and cognitive outcome (Full Scale IQ, grey bars) for both the perinatal arterial ischaemic stroke (PAIS) and the periventricular haemorrhagic infarction (PVHI) group.

Individual data indicated that nine children with PVHI (n=21) showed a significant decline (greater than expected based on the SEM) in intellectual ability from early developmental assessment to cognitive assessment at school age. Seven children showed a significant improvement in intellectual ability. In the PAIS group (n=29), 16 children showed a significant decline and three children showed significant improvement in overall intellectual ability over time.

Predictors of intellectual outcome at school age

Among the PVHI group, univariate analyses showed lower FSIQ scores in children with PHVD (80 vs 96; 95% CI mean difference 2–31). As the level of maternal education was not normally distributed, the correlation between FSIQ and maternal educational level was tested non-parametrically, showing a positive correlation (r=0.40; p=0.005). No effect on FSIQ was found for location (frontal, parietal, or frontoparietal) and side (left vs right) of the lesion, handedness, sex, gestational age, or development of postneonatal epilepsy and unilateral spastic CP.

In the PAIS group, univariate analyses showed lower mean FSIQ scores in children with a main branch middle cerebral artery stroke (75 vs 92; 95% CI mean difference 3–31), in children with involvement of the basal ganglia and thalami (80 vs 96, 95% CI mean difference 4–29), in children who developed postneonatal epilepsy (74 vs 91, 95% CI mean difference 3–32), and in children who developed unilateral spastic CP (78 vs 95, 95% CI mean difference 5–29). Multivariate analyses showed that involvement of the basal ganglia and thalami, and development of postneonatal epilepsy were the best predictors for FSIQ (Table 3). No effect on FSIQ was observed for side of the stroke, gestational age, sex, involvement of the cortex, and level of maternal education.

Table 3. Multiple linear regression results for factors associated with Full-scale IQ in the perinatal arterial ischaemic stroke group
 BSE Bβ
  1. R2=0.21 for step 1. R2=0.33 for step 2.

Step 1
Constant96.154.48 
Basal ganglia and thalami involvement−16.286.03−0.46
Step 2
Constant98.364.33 
Basal ganglia and thalami involvement−14.005.77−0.40
Postneonatal epilepsy−14.366.70−0.35

Discussion

In this study, we have shown that both children with PVHI and children with PAIS have below average cognitive ability at school age and some weaknesses in other specific cognitive domains. While a neurodevelopmental delay was already present in infancy in participants with PVHI, children with PAIS showed a significant decline in neurodevelopmental outcome over time.

Periventricular haemorrhagic infarction

Children with PVHI already showed performance outside the normal range in infancy and they demonstrated no further decline when tested at school age. At school age, children with PVHI showed an average FSIQ of 86. While this was below that of typically developing children, one should take into account that the cognitive performance of these children might be further complicated by their preterm birth, which is a known risk factor for lower IQ scores. Several studies have shown that IQ scores of preterm-born children are 5 to 10 points lower than in a term-born comparison group.[20] Children with PVHI, therefore, seem to perform only slightly worse than preterm children without PVHI. This is in line with the work of Roze et al.,[7] on cognitive function in infants with PVHI, who found a mean FSIQ of 83 in these children. Bassan et al.[21] also found that half of PVHI survivors developed significant cognitive abnormalities.

Although previous studies reported impairments in verbal memory[7, 22] and behavioural problems (especially impairments in executive functions),[22, 23] we could not confirm this. Preterm birth is known to affect visuomotor processes,[24, 25] which is in line with our findings. Similarly, attention and concentration deficits are more frequently observed in preterm children.[26, 27] It is unknown, however, whether these deficits are exacerbated by the presence of a PVHI. Finally, we found shorter reaction times in response to visual stimuli in children with PVHI. We hypothesize that the increased use of electronic devices may play a role in this finding.

Although we were not able to perform a linear regression analysis in the PVHI group, our data suggest that development of PHVD had a negative effect on cognitive performance. Posthaemorrhagic ventricular dilatation is a frequently observed complication of an intraventricular haemorrhage. Although it is not necessarily associated with an unfavourable outcome, even when neurosurgical intervention is required, several studies have shown that PHVD in the presence of PVHI results in lower cognitive performance.[7, 28]

The observed relation between the level of maternal education and FSIQ has been reported before.[29, 30] Possible explanations include a higher quality of child care, cognitive stimulation, and the use of follow-up services. Moreover, IQ differences in children may be genetically caused by differences in the maternal IQ.[29]

Perinatal arterial ischaemic stroke

The reported cognitive outcome after PAIS has been variable. Some studies have reported a favourable long-term outcome,[31] whereas other studies have reported lower intellectual abilities at school age.[8, 32-34] Ricci et al.[6] found normal cognitive outcome in children with PAIS, but the majority of the participants were preschool children at the time of assessment. Ballantyne et al.[31] reported longitudinal measurements and found a normal cognitive function at school age, but found lower IQ scores in children who developed epilepsy. The decline in function over time, as found in our study, is in line with the study of Westmacott et al.,[8] who found that the performance of children with PAIS did not differ from the normative sample of preschool children, but when compared with the normative sample of school-age children, performance was significantly lower for overall intellectual ability. This does not reflect a loss of skills, but, rather, slower gains over time and difficulty with higher-level cognitive skills compared with healthy peers. The reported normal performance in other studies might reflect differences in the duration of follow-up, which may have been too short to detect late-emerging deficits. This suggests that a continued follow-up of these children until they are attending school is important to understand how these children cope with more academic tasks.

Previous studies have reported specific language impairments after PAIS.[31, 35] In contrast to these studies, we did not find impairments in word comprehension, but we did find poor visual–motor integration. Talib et al.[36] did not find impairments in visual–motor integration, but this may also be attributed to the short duration of their follow-up.

Linear regression showed that both involvement of the basal ganglia and thalami and development of postneonatal epilepsy were associated with lower FSIQ scores. Development of postneonatal epilepsy has previously been associated with a decline in cognitive function.[31] It remains uncertain, however, what the underlying mechanism is, as both epilepsy and the antiepileptic drugs may play a role. We were unable to differentiate between these two factors, as all children with epilepsy used antiepileptic drugs. Thalamic injury is increasingly being associated with poor cognitive outcome, in both preterm and term infants, possibly owing to the regulatory role of the thalamus in information transmission to the cortex and between cortical areas.[37-39] It has been extensively associated with cognitive deficits in adult stroke,[40] but ours is the first study that associates thalamic injury with lower cognitive scores after PAIS.

This study has some limitations which need to be addressed. First, the comparison of performance across time was complicated by the use of a neurodevelopmental measurement and an intelligence test at school age. After discharge from the neonatal intensive care unit, all children were seen at the neonatal clinic for routine follow-up. In this setting, the Griffiths Mental Developmental Scale provided a useful tool to evaluate the child's development in a limited amount of time. A second potential limitation is that, although we studied a relatively large group with perinatal brain injury, a larger group would allow more extensive statistical analyses, potentially identifying other factors associated with outcome. Another possible limitation is that the age range at neuropsychological assessment was very wide (6–20y). We controlled for this using age-appropriate norms and test versions. All the children in our study were attending school, but it is important to understand how they cope with more academic tasks. Finally, we did not study the visual performance of all children, though prematurity and both PAIS and PVHI are associated with visual disorders, including retinopathy of prematurity, visual field defects, and cerebral visual impairment. Even though visual disorders may also affect the neuropsychological test results, the presence of such disorders seemed low in the present study, as a parental questionnaire reported visual deficits in only three children (one quadrantanopia, two hemianopia) and screening for retinopathy of prematurity in 22 out of 28 preterm-born infants displayed a grade I retinopathy in only one infant.

In conclusion, children with PVHI or PAIS reported in this study showed a below average performance when tested at school age. Their performance was, however, not severely affected, as it fell within 1SD of the healthy population and a majority (76%) still attended mainstream education. While motor outcome can be reliably predicted using conventional neonatal MRI, prediction of cognitive outcome remains challenging, as only development of PHVD after PVHI, and basal ganglia and thalami involvement and development of postneonatal epilepsy after PAIS, were associated with poorer cognitive outcome.

Acknowledgements

Niek van der Aa was funded by grants of the Wilhelmina Children's Hospital Research Fund and of the Dutch Phelps Foundation (www.phelps-stichting.nl).

Ancillary