Neurodevelopment After Perinatal Arterial Ischemic Stroke

The neonates in this study comprised 2 cohorts of term newborn infants who were admitted to the NICU or referred for neurologic assessment to Queen Charlotte’s or Hammersmith Hospital in London, United Kingdom (n = 79) or the Wilhelmina Children’s Hospital of the University Medical Center in Utrecht (UMCU), the Netherlands (n = 117) between October 1990 and January 2015. All infants had acute symptoms in the first week after birth, most often (hemi) convulsions, but in a few infants their symptoms were less neurologically specific. All had PAIS confirmed on their neonatal MRI. Eight infants were excluded because of congenital syndromes with known adverse outcome (n = 4) or other significant brain lesions (n = 4), resulting in a total cohort of n = 187 infants. Infants who died in the neonatal period (n = 3), whose neonatal MRI scan could not be reevaluated (n = 4), or who were lost to follow-up (n = 20) were excluded from further analyses (Supplemental Table 7). This resulted in a total study cohort of 161 term neonates.

Informed verbal parental consent was obtained to perform an MRI for clinical purposes. The Institutional Review Board of the UMCU approved the use of MRI data for anonymous data analysis and waived the requirement to obtain written informed consent. In London, neonatal MRI scans were performed after obtaining written informed consent and permission to use these scans and clinical data for research.

In both centers, MRI was performed on a 1.0, 1.5, or 3T whole-body system (Philips Medical Systems, Best, Netherlands, or Picker System, Cleveland, OH) by using a scanning protocol including at least T1-weighted, T2-weighted, and diffusion-weighted imaging (DWI). In general, infants were sedated to minimize movement artifacts. Because we included infants over a long time period, the MRI protocol was not always the same; imaging details have been reported previously.^6,9,16 

Neonatologists experienced in neonatal brain MRI (F.C. and L.dV.) reevaluated each (of both centers) MRI scan. The lesions were assessed in 3 planes, if possible. On the basis of shape, extent, and localization of the area of signal intensity changes, all infants were classified to 1 of the stroke subtypes shown in Fig 1 on the basis of their most predominant stroke pattern.

Classification was mostly based on vascular territory of specific named arteries that resulted in characteristic infarctions as described by Govaert.¹⁵ However, we felt that consistency of involvement of particular anatomic structures was more important, and the involvement of specific anatomic hallmarks was also used for classification. A hemispheric lesion in the MCA territory located posterior to the central sulcus was attributed to the posterior branch of the MCA, whereas involvement anterior to the central sulcus was attributed to the anterior branch. When the full central sulcus was involved, but not regions more anterior or posterior, this was attributed to the middle branch MCA. If the anterior or posterior areas and middle branch MCA were involved, the most predominant branch was chosen, and central sulcus involvement was noted separately. Involvement of the central sulcus region was best assessed in the parasagittal plane. Where >1 MCA branch artery was involved this was also noted separately as multiple lesions. Small punctate lesions were not considered part of the spectrum of “multiple lesions.”

Magnetic resonance images were also evaluated for involvement of specific regions we considered likely to be of major importance in predicting outcome (ie, the CST, the central sulcus region, thalami, and basal ganglia, as described previously^2,8), mainly by visual inspection of the DWI from MRI scans done in the MRI from the first week and from T1- and T2-weighted images when the MRI was acquired later.⁷ Involvement of the CSTs has been described previously.¹¹ Only when the middle part of PLIC and cerebral peduncle were affected, carrying the main motor tracts, was this classified as involvement (Fig 2) for the purposes of this study.

Often with hemispheric strokes, signal changes were also seen in the thalami, particularly the pulvinar, that are likely secondary to “network” injury rather than part of the primary stroke.¹⁷ Examples of primary thalamic stroke (perforator stroke) and secondary network injury to the thalamus are shown in Fig 1 (anterior MCA). Bilateral lesions were described as bilateral stroke when stroke lesions were equally severe or as smaller contralateral lesions when 1 region of stroke predominated.

Neurodevelopmental outcome was determined during routine follow-up appointments. We only used data from after 12 months until 7 years.

Cognitive development was determined by using the developmental quotient of the Griffiths Mental Development Scale (GMDS), calculated by using all subscale scores except locomotion, the Bayley Scales of Infant and Toddler Development, Third Edition (BSITD-III), or the Wechsler Preschool and Primary Scale of Intelligence second or third edition.^18,–20 For all cognitive tests, z scores were calculated to allow comparison of the data for statistical analyses. Cognitive delay was defined as a z score below −1, corresponding to −1 SD. Language delay was defined as a language score on the GMDS of <−1 SD, >15 points on the Dutch language screening instrument,²¹ or a diagnosis of speech and/or language disorders.

CP was diagnosed by using criteria from the European CP Network,²² and severity was determined by using the Gross Motor Function Classification System.²³ Behavioral problems were asked about at each clinic visit, and, when appropriate, children were referred to a clinical psychologist; additionally, in the UMCU cohort, parents completed the Child Behavior Checklist.²⁴ Only behavioral assessments after 2 years of age were taken into account. Postneonatal epilepsy was classified as (recurrence of) seizures diagnosed on EEG for which regular medication was given. Visual field defects included hemianopia and quadrantanopia, diagnosed by a specialized visual development unit in London or pediatric ophthalmologist specialized in assessing visual field defects in infancy in Utrecht. An adverse outcome was defined as the presence of 1 or more of the following: CP, cognitive deficit, language delay, epilepsy, behavioral problems, or visual field defect at latest follow-up.

Descriptive statistics are summarized as percentages of the available study cohort or as median and interquartile range (IQR) where appropriate. Stroke subtypes and other imaging features were compared with outcome parameters by using χ² tests, independent t tests, or Mann–Whitney U tests (for nonparametric variables). Binary logistic regression analysis was performed to determine independent MRI predictors for adverse outcome, which were expressed as odds ratios (ORs) with 95% confidence intervals (CIs). These regression analyses were performed for the total cohort and also separately for those without main MCA branch infarction.

Statistical analyses were performed with SPSS, version 21 (IBM SPSS Statistics, IBM Corporation). P values <.05 were considered to be statistically significant.

The total study cohort consisted of 161 term neonates born at a median of 40.3 weeks’ gestation and a median birth weight of 3440 g (Table 1). There were no differences in infant clinical parameters between the 2 centers, so results are reported for the total cohort. In Utrecht, 30 infants received erythropoietin as part of an intervention study or off-label use.²⁵ These infants did not differ in stroke patterns and also had equal rates of adverse outcome (unpublished data N.W., N.A., F.G., M.B. and L.dV.) and were therefore not reported separately.

MRI was performed at a median of 5 (IQR: 3.5–7) days after birth. The MCA was most commonly involved, and most often affected were the main branch, posterior, and perforator branches (19%, 17%, and 17%, respectively) and less often the anterior, middle, or 1 of the cortical branches (11%, 13%, and 13%, respectively). PCA stroke was found in 14 cases (9%) and ACA stroke in 2 cases (1%). Clinical characteristics are given in Table 1 and details of involvement of specific regions in Table 2. Involvement of the CST could not be determined in 5 infants (3%), who had an MRI in the second week after birth without DWI abnormalities and T1- or T2-weighted imaging of insufficient quality to assess the CST. Infants that were lost to follow-up (Supplemental Table 7) did not differ in terms of stroke pattern classification, basal ganglia and thalami (BGT), or CST involvement.

Infants with cortical MCA infarction were born at significantly later gestational age (mean 40.7 ± 1.3 vs 40.0 ± 1.4 weeks; P < .03) compared with other subtypes.

Hypoglycemia was more often found in infants with PCA infarction compared with the other subtypes (58% vs 16%; P < .0001) and more often in main branch MCA strokes compared with other subtypes (36% vs 16%; P < .02), whereas the incidence of hypoglycemia was not significantly different between PCA and main branch MCA stroke (P > .1).

Seizures at presentation were less common in perforator stroke compared with the other subtypes (62% vs 94%; P < .0001). Also, perforator stroke was less often left sided compared with the other subtypes (48% vs 71%; P < .03). Other characteristics from Table 1 were not significantly different between stroke subtypes.

All infants were seen between 12 months and 7 years with a median age of 41.7 months when last seen (Table 3). There were no differences in outcomes between centers. At their latest follow-up, 49 infants (30%) had developed CP; Gross Motor Function Classification System levels were determined for 40 infants with 90% at level I, 8% at level II, and 3% at level IV (1 child with bilateral main branch MCA). There were no infants in this study who were first diagnosed with CP beyond 2 years of age. Cognitive test results were available for 157 infants, and 3 infants could not be tested because of severe delay (<−2 SD). This resulted in 37 infants (23%) with a cognitive z score <−1 SD and 13 infants (8%) with a cognitive z score <−2 SD (Table 3). At their latest follow-up, 87 of 161 infants (54%) had an adverse outcome; 38 infants (24%) had 1 or a combination of adverse outcomes without having CP. More details on adverse outcome domains per stroke subtype are given in Table 4. Further analyses were performed by using outcome results from each patient’s last follow-up.

Overall, 50 of 87 (57%) infants with adverse outcome developed sequelae in multiple domains. Of the 49 infants with CP, 35 (71%) had another adverse outcome, most commonly a cognitive deficit (n = 22; 45%). Visual field defects did not occur in isolation, and most often, they occurred in combination with CP (13 of 17). Although adverse outcomes commonly co-occurred, only 4 children were affected in all 6 developmental domains.

Infants with language delay had increased risk of cognitive delay (OR: 11.8; 95% CI 4.7–29.2), but excluding those with main MCA branch stroke, the odds for cognitive delay were 6.5 times increased with language delay (95% CI 2.1–20.1). Postneonatal epilepsy also increased the risk for cognitive delay (OR: 9.1; 95% CI 3.1–26.6), but this was only significant in the main MCA branch stroke group.

Analyzing all infarcts together, univariate analyses revealed associations between several MRI parameters and neurodevelopmental outcome domains (Table 5). Involvement of the cerebral peduncles and combined involvement of the BGT were both related to almost all adverse outcome domains with ORs ranging between 3.8 and 115.6 (Table 5).

For the total cohort, several MRI parameters were significantly and independently associated with different outcome domains, as described in Table 6. Because there was involvement of the CST, basal ganglia, and central sulcus in all main MCA branch infarcts, multivariable analyses were repeated separating all infants with main MCA branch infarction from the others (Table 6).

In infants with main MCA branch infarction, no specific MRI parameters were associated with different adverse outcome domains. In the other subgroup (n = 130), CP was still associated with involvement of the cerebral peduncle (OR: 34.4; 95% CI 5.6–208.9) and combined BGT involvement (OR: 6.9; 95% CI 1.1–45.1). Adverse cognitive outcome was associated with combined BGT involvement (OR: 4.4; 95% CI 1.2–16.9). Language delay and visual field defects were no longer associated with MRI features, but postneonatal epilepsy remained associated with involvement of the cerebral peduncle (OR: 9.7; 95% CI 1.0–101.1) and bilateral lesions (OR: 15.9; 95% CI 2.3–110.8). Behavioral problems were associated with combined BGT involvement (OR: 8.9; 95% CI 1.9–41.2). Combined involvement of the BGT increased the risk of adverse outcome in at least 1 domain (OR: 17.1; 95% CI 2.1–141.1; Table 6; Fig 3).

In this study, we have demonstrated that an adverse outcome in term infants with PAIS depends on stroke territory (the site, extent, and location of the lesion). To the best of our knowledge, this is the first study in which is provided a precise overview of a spectrum of different outcome domains per stroke pattern in a large cohort of term infants from 2 centers. Additionally, with our study, we highlight the use of early neonatal MRI, and especially DWI, to predict neurodevelopmental outcome, because outcome is not only dependent on stroke territory but also on the involvement of specific brain regions, such as the CST.

It is of great importance to clinicians and parents to evaluate promptly the risk of an adverse outcome in patients with PAIS, because early intervention strategies, which may attenuate unfavorable development, need to be appropriately focused.²⁶ It is also important when possible to reassure parents that outcomes are likely to be good. Several studies on neurodevelopment after PAIS are available, but authors of these studies report a wide range of abnormal neurodevelopment, mainly because there is no distinction between specific stroke subtypes.^27,28 This makes risk estimation difficult for the individual child. With our study, we have distinguished several specific PAIS subtypes and described incidence rates per outcome domain for them, enabling more personalized prediction of long-term development. In the literature, the incidence rate of CP after PAIS is ∼30%, comparable to our study.^6,8,27,28 However, infants with main MCA branch infarction will all develop CP, whereas this percentage ranges between 0% and 21% in other stroke subtypes. With our data, we provide a firm basis for informing parents of infants with main MCA branch infarction differently about future prospects than parents of infants with other stroke subtypes.

Involvement of the CST on neonatal MRI was seen in the majority of patients with more extensive stroke subtypes, as reported in other studies.^8,29 The CST signal changes are best seen on early DWI and have been described as “pre-Wallerian degeneration.” When they are seen in the middle third of the cerebral peduncle they are always associated with the development of hemiplegic CP.^7,9 Because pre-Wallerian degeneration is the result of anterograde degeneration of the descending axons of injured cell bodies within the infarcted areas, it was found more often in infants with larger infarctions (affecting the complete motor cortex). Infants with other stroke subtypes not resulting in involvement of the CST in our cohort did not develop CP. We performed multivariable modeling separately for those without main MCA branch infarction, to determine individual MRI risk factors for adverse outcome in milder subtypes. For these subtypes, involvement of the cerebral peduncle was still a risk factor for CP and epilepsy, illustrating the importance of pre-Wallerian degeneration in the prediction of adverse outcome in those with less widespread stroke.

Involvement of the BGT increased the risk for CP, in agreement with the literature.^8,30,31 In our cohort, BGT involvement most commonly occurred with larger infarcts (main and partial MCA branches) and was most often part of the primary stroke. But BGT involvement could also be a manifestation of secondary injury to connectivity pathways (eg, corticothalamic or corticostriatal networks) particularly in the thalami and best seen on DWI.^17,30,32 The increased risk for CP most likely stemmed from the larger stroke than just the BGT involvement. This is supported by several studies revealing that primary stroke lesions restricted to the BGT (ie, perforator stroke) are usually not associated with adverse motor outcome.^6,33 In our cohort, only 15% of infants with perforator stroke developed unilateral spastic CP, all related to additional involvement of the PLIC.

Authors of a limited number of studies reported on long-term cognitive outcome in infants with PAIS.^27,34,–36 We found that cognitive delay occurred in 23% of all PAIS patients but in 57% of those with a main branch MCA stroke when last seen at a median age of 41 months. Authors of other studies have reported even higher rates of cognitive impairment after PAIS at school age.^36,37 Because most infants were still young when last seen, we did not see a trend over time. Multivariable analysis revealed that cognitive delay was related to BGT involvement. However, this seemed to reflect larger strokes, because BGT involvement was most often seen in main and partial branch MCA strokes. Other studies have revealed that larger infarct volume was associated with adverse cognitive development in PAIS.^38,–41 We also found that posterior MCA branch and PCA strokes had higher rates of cognitive delay compared with other nonmain MCA subtypes, indicating that not only volume but also location of affected tissue plays an important role in cognitive development. In a recent study, Stephan-Otto et al⁴² reported that stroke in regions posterior to the central sulcus, close to the arcuate fasciculus, may account for language deficits after PAIS. In our cohort, rates of language delay were not higher in posterior compared with anterior MCA branch strokes, but many children were too young for detailed speech or language assessment. It was of interest that infants with language delay had 6.5 to 10 times increased risk of cognitive delay, revealing that language and cognition are closely related. However, cognitive delay might also precede language delay or share a common origin, and exact causative mechanisms need to be studied further. Development of postneonatal epilepsy increased the risk for cognitive delay as described previously, but this was limited to infants with main MCA stroke.⁴³ 

This study has several limitations inherent to its retrospective design. Infants were only eligible if they were admitted to the NICU or referred for neurologic assessment, excluding infants with (smaller) infarcts that may not have caused neonatal symptoms; also, preterm infants were not included.⁴⁴ However, our strict inclusion criteria resulted in a homogeneous group of term infants with PAIS. This study was focused on DWI from MRI, whereas early DWI may not always be possible in all institutions. However, we were often able to see signal intensity changes in CST on T2-weighted sequence as well, especially when the MRI was done in the second half of the first week. The use of apparent diffusion coefficient (ADC) maps from DWI to assess acute ischemic injury is recommended to avoid T2 shine-through and other artifacts. Because the ADC map was not always available for our cohort, we used DWI for all infants. When ADC maps were available, they were used to verify DWI signal abnormalities, as is recommended in clinical practice. Infants that performed well were sometimes discharged from follow-up, and cognitive, language, and behavioral problems may have been underdiagnosed.^36,37 Long-term outcome studies with a prospective design are needed to determine whether early predictions in PAIS patients remain stable over time. Even with this large study, some subgroups were small, and we cannot exclude that some associations between brain regions and outcomes might have been significant.

In a large cohort of term-born infants from 2 centers, we have demonstrated that neurodevelopmental outcomes vary between PAIS subtypes. Although neurodevelopmental outcome was invariably adverse in at least 1 domain with main branch MCA stroke, in other PAIS subtypes, outcome was normal in 43% to 71% of children. With this study, we provide clinicians with important information for more precise risk evaluation of neurodevelopment in PAIS patients on the basis of the tissue involved, assessed from early MRI allowing better counseling of parents in the neonatal period, personalized planning of therapeutic interventions, and long-term support for behavioral and cognitive difficulties.

ACA

anterior cerebral artery

ADC

apparent diffusion coefficient

BGT

basal ganglia and thalami

BSITD-III

Bayley Scales of Infant and Toddler Development, Third Edition

CI

confidence interval

CP

cerebral palsy

CST

corticospinal tract

DWI

diffusion-weighted imaging

GMDS

Griffiths Mental Development Scale

IQR

interquartile range

MCA

middle cerebral artery

OR

odds ratio

PAIS

perinatal arterial ischemic stroke

PCA

posterior cerebral artery

PLIC

posterior limb of the internal capsule

UMCU

University Medical Center in Utrecht