Abstract
Objective
We sought to determine the clinical features of hemiplegic cerebral palsy associated with perinatal arterial ischemic stroke or periventricular venous infarction.
Methods
We studied children with hemiplegic cerebral palsy enrolled at 9 rehabilitation centers across Ontario. We compared children with underlying perinatal arterial ischemic stroke or periventricular venous infarction on clinically acquired brain imaging. Analysis also included prenatal (maternal, prenatal/gestational) and perinatal (obstetrical, neonatal) clinical features collected from birth records and standardized parent interviews.
Results
The 144 children with hemiplegic cerebral palsy (62% male) included 95 with perinatal arterial ischemic stroke and 49 with periventricular venous infarction. In this cohort of children with hemiplegic cerebral palsy, we found neonatal systemic thrombosis (ie, blood clots in the body) (P = .05), emergency cesarean section (P = .05), and neonatal seizures (P = .01) to be clinical features associated with hemiplegic cerebral palsy in children with perinatal arterial ischemic stroke more often than periventricular venous infarction. Preterm delivery rates were similar for perinatal arterial ischemic stroke and periventricular venous infarction.
Conclusion
We determined clinical features associated with the 2 most typical forms of focal ischemic brain injury in children with hemiplegic cerebral palsy, including mode of delivery emergency cesarean section, neonatal seizures and systemic thrombosis. These findings provide further insight and support for existing findings about focal brain injury patterns leading to hemiplegic cerebral palsy in children.
Cerebral palsy is the most common physical disability in early childhood. 1 Hemiplegic cerebral palsy, characterized by motor impairment on one side of the body, is associated with unilateral or asymmetric brain lesions acquired at different stages of brain development during gestation and the perinatal period. Hemiplegic cerebral palsy, the most common subtype of cerebral palsy, accounts for 38% of cerebral palsy cases. 2 The prevalence of hemiplegic cerebral palsy provides a compelling reason to focus efforts on prevention. Prevention of vascular brain injuries associated with hemiplegic cerebral palsy requires, in turn, an understanding of the underlying pathogenesis. Differentiating the patterns of brain injury is a crucial first step. Determining the clinical features and patterns of pre- and perinatal risk factors associated with distinct patterns of brain injury may provide a deeper understanding of preventable causes of hemiplegic cerebral palsy. The focus of the current article is to differentiate the 2 most common patterns of brain injury in hemiplegic cerebral palsy, arterial ischemic stroke, or periventricular venous infarction, according to their associated clinical features in an Ontario-based population study, the Hemiplegic Cerebral Palsy Network (Hemi CP-NET).
Multiple magnetic resonance imaging (MRI) classification systems have been published for all types of cerebral palsy.2,3–5 In children with hemiplegic cerebral palsy, more than half are associated with perinatal stroke, and more than one-third with white matter injury of prematurity.2,6,7 Perinatal stroke, the most common brain lesion underlying hemiplegic cerebral palsy, results from focal disruption of cerebral blood flow by focal hemorrhage or isolated vascular occlusion. The timing of perinatal brain injury ranges from 20 weeks of fetal life through the 28th postnatal day. 8 Perinatal arterial ischemic stroke is a focal ischemic infarct resulting from cerebral arterial occlusion with a clot, most commonly of embolic origin, from the heart, systemic venous system, or the placenta. Perinatal arterial ischemic stroke infarcts occur in stereotyped patterns representing the major arterial territories, including the anterior, middle, or posterior cerebral artery and involving the cortex, subjacent white matter, or the thalamus/basal ganglia. Acutely diagnosed perinatal arterial ischemic stroke leading to hemiplegic cerebral palsy is typically identified in newborn infants with seizures,9,10 some of whom develop hemiparesis. However, others present with emerging hemiparesis in later infancy, with associated brain imaging demonstrating a remote perinatal arterial ischemic stroke. This entity is classified as “Presumed perinatal arterial ischemic stroke (PAIS).” 11 Research on clinical features associated with subtypes of brain lesions involved in hemiplegic cerebral palsy, including a large, case-control study, found peripartum clinical factors associated with perinatal arterial ischemic stroke, 12 confirming earlier reports. 13 However, these studies excluded preterm-born children.
Periventricular venous infarction is another form of “presumed perinatal ischemic stroke” consisting of a focal venous infarct in the periventricular white matter. The periventricular venous infarction appears as a stereotyped, unilateral, fan-shaped pattern of focal encephalomalacia. It results from compression of venous outflow by the medullary or terminal veins in the subependymal zone, typically occurring in the premature brain. 14 Periventricular venous infarction is a subtype of white matter injury diagnosed either as “acute periventricular venous infarction” diagnosed in premature infants with intraventricular or germinal matrix hemorrhage 14 and secondary venous infarction of the periventricular white matter 15 or as “remote periventricular venous infarction” detected when the initial brain imaging is performed later in infancy. 11
In an earlier Hemi CP-NET study, Fehlings et al 7 described a novel neuroimaging classification system explicitly designed to characterize the brain lesions associated with clinical hemiplegic cerebral palsy on CT and MRI. Using this imaging classification, we sought to determine the pre- and perinatal clinical features associated with either perinatal arterial ischemic stroke or periventricular venous infarction in our well-characterized geographically defined cohort of children with hemiplegic cerebral palsy.
Methods
Study Design
Cross-sectional sampling for Hemi CP-NET was conducted across 9 clinical centers throughout Ontario, Canada, affiliated with the Cerebral Palsy Integrated Neuroscience Discovery Network (CP-NET). These included Holland Bloorview Kids Rehabilitation Hospital, McMaster Children's Hospital, the Hospital for Sick Children, Grandview Children's Centre, Erin Oak Kids Centre, London Health Sciences Centre, Hotel Dieu Hospital, Ottawa Children's Treatment Centre, and the Health Sciences North Centre. The study occurred between April 2012 and October 2017. Research Ethics Board approval and written informed consent were obtained.
Participants
Children with hemiplegic cerebral palsy were identified during outpatient visits at study centers, supplemented by screening health records for the diagnosis of hemiplegic cerebral palsy. Eligibility criteria included (1) 2-18 years of age at enrolment, (2) a diagnosis of unilateral or hemiplegic cerebral palsy, (3) brain imaging studies available, including magnetic resonance imaging (MRI) or computed axial tomography and (4) clinical data available on pre- and perinatal clinical features.
Data Collection
Clinical data were collected from health records, including pre- and perinatal records at the enrolling center, records obtained from the birth hospital, and standardized parent interviews conducted by physicians (developmental pediatricians) and occupational therapists at in-person study visits (previously described). 7 Clinical features/markers identified by health chart review and parental report were classified into 4 categories: neonatal, maternal, pregnancy-related, and labor and delivery-related. Gestational age at birth was classified by the number of gestational weeks when available and separately as “term” vs “preterm” (defined as <37 weeks’ gestation) delivery. Clinically acquired brain imaging studies, including MRI, computed tomography (CT), and cranial ultrasonography, were obtained from the Ontario hospital where the imaging was conducted. These images were sent by mailed computerized disc or electronic file transfer to the Stroke Imaging Lab for Children (based at the Hospital for Sick Children, Toronto, Canada) and transferred into the research imaging database. When multiple neuroimaging studies were available for individual children, all were reviewed before completing the neuroimaging classification form.
Neuroimaging Analysis
Brain imaging was centrally reviewed by the study neuroradiologists (PK, MM, MS) using the “Hemi CP-NET” study Neuroimaging scoring form (previously described 7 ). Brain lesions were classified by lesion patterns including subtypes: (a) perinatal arterial ischemic stroke including middle cerebral artery or non–middle cerebral artery patterns of arterial infarction; (b) periventricular/ventricular patterns including periventricular venous infarction or nonperiventricular venous infarction white matter injury (other forms of periventricular white matter injury), and intraventricular hemorrhage was identified and classified by subtypes I-IV; (c) unilateral brain malformations; and (d) miscellaneous (for complete form see Fehlings 7 Appendix S1 “Neuroimaging Data Collection Form”). Criteria for lesion pattern classification are based on previously established definitions.4,14–16 Perinatal arterial ischemic stroke is defined as a focal area of brain infarction conforming to a known cerebral arterial territory. Periventricular venous infarction is defined as unilateral periventricular focal encephalomalacia, presumed sequelae as obstruction of the deep medullary veins, with relative sparing of the overlying cortex and basal ganglia. 7 In addition to the lesion patterns, lesions were also classified by additional characteristics including (1) cerebral arterial territory of infarct using main artery and subbranch labels for perinatal arterial ischemic stroke (eg, middle cerebral artery–posterior division), (2) type of tissue affected (cortical or deep gray matter, white matter), (3) anatomic structures/substructures affected (n = 15: frontal, parietal, temporal or occipital lobes, basal ganglia substructures putamen caudate globus pallidus, thalamus, cerebellum, brainstem and others), and (4) eloquent gray matter and white matter tracts affected (n = 7). Interrater reliability was assessed as previously reported 7 on the initial 85 cases, which were double-scored by 2 neuroradiologists (MM, PK). Initial agreement averaged 93.6% (range 74%-99%) across all lesion characteristic scoring. 7
Statistical Analysis
The continuous variable age at imaging was assessed using a t test, with descriptive statistics calculated and summarized with means and SDs. For categorical variables, Fisher exact test and the associated odds ratios or nonparametric independent samples Wilcoxon-Mann-Whitney U tests were conducted to explore univariate relationships between perinatal arterial ischemic stroke and periventricular venous infarction groups. Following univariate analysis, variables with a significance level of P <.05 were fit using Firth's logistic regression models 17 to mitigate small sample size bias because of a small data set or rare outcomes that create challenges when analyzing binary outcomes. This model also corrects the predicted probabilities by a post hoc adjustment of the intercept. 18 Odds ratios (ORs) were estimated for the clinical features or markers associated with the perinatal arterial ischemic stroke or periventricular venous infarction patterns of brain injury. Because of potential bias caused by missing data, imputation approaches were explored for variables with at least 20% missing data in either the periventricular venous infarction or perinatal arterial ischemic stroke group, and variables with 20% or more missing data were excluded from multivariate models. In the multivariate model, odds ratios, 95% confidence intervals, and P values are presented.
Results
Participant Characteristics
We identified 320 children with hemiplegic cerebral palsy. Of these, 75 with no clinical neuroimaging available were excluded from this analysis. Among the remaining 245 children, 106 had periventricular white matter patterns of injury, including 49 with periventricular venous infarction. Another 100 children had perinatal arterial ischemic stroke (97 middle cerebral artery and 3 non–middle cerebral artery territories), and there were 19 with unilateral brain malformations, 12 with miscellaneous abnormalities, and 8 with normal scans. Five children with middle cerebral artery–perinatal arterial ischemic stroke were excluded because of missing extensive clinical features/markers data. The final study cohort included 144 children, 95 with perinatal arterial ischemic stroke and 49 with periventricular venous infarction (see Table 1). Male sex comprised 61.8% of the entire group, 61% in perinatal arterial ischemic stroke and 63% in periventricular venous infarction. Age at hemiplegic cerebral palsy diagnosis was, on average, 1.29 years and was older in the periventricular venous infarction group (22.4 months vs 11.89 months in the perinatal arterial ischemic stroke group, P = .006). Ethnicity was predominately White in both groups (70% in perinatal arterial ischemic stroke and 73% in periventricular venous infarction). Clinical manifestations of hemiplegic cerebral palsy, detailed in our prior analysis, included increased distal tone in the affected arm and leg in >80%. 7 Children with perinatal arterial ischemic stroke had decreased grip strength on the hemiplegic side more frequently than periventricular venous infarction (65.9% vs 88.2%, respectively) and decreased full-scale IQ (35.5% vs 53%, respectively).7
Demographics and Clinical Characteristics of Children With Hemiplegic Cerebral Palsy Due to PAIS or PVI.
Abbreviations: C-section, cesarean section; CT, computed tomography; MRI, magnetic resonance imaging; PAIS, perinatal arterial ischemic stroke; PVI, periventricular venous infarction.
In postnatal months.
At or below P = .05.
Infertility treatment (any) included hormone medication, intrauterine insemination, and in vitro fertilization. In vitro fertilization was reported in 50% of both groups: 4/8 in the PVIV and 1/ 2 PAIS groups.
Meconium was not evaluated statistically because of >20% missing data.
Prenatal and Perinatal Clinical Features Associated With the Perinatal Arterial Ischemic Stroke and Periventricular Venous Infarction Subgroups
Prenatal and perinatal characteristics of children with perinatal arterial ischemic stroke and periventricular venous infarction are summarized in Table 1. Although univariate analysis revealed significant differences between perinatal arterial ischemic stroke and periventricular venous infarction in mean gestational age (x̄ = 38.5 ± 2.51 and 37.2 ± 4.17, respectively, P = .04), the proportion born at term (37 to <43 weeks’ gestation) vs preterm (<37 weeks) was similar (x̄ = 14 [14.7%], x̄ = 9 [18.4%]) (Figure 1). Additional variables showing significant differences between perinatal arterial ischemic stroke and periventricular venous infarction were neonatal seizures at birth (perinatal arterial ischemic stroke n = 28 [29.5%] and periventricular venous infarction n = 1 [2.0%], respectively, P = .05), neonatal systemic thrombosis (7.4% and 0%, P = .05), infertility treatment (2.1% and 16.3%, P = .002), and emergency cesarean section (37.9% and 20.4%, P = .02). These variables were entered into multivariate analysis. Four additional variables approached or were statistically significant (P < .02), but were excluded from multivariate analysis because of missing >20% of data. These included a family history of thrombosis, intrapartum asphyxia or hypoxic-ischemic encephalopathy, congenital malformation or syndrome, meconium, and Apgar score at 5 minutes. Multivariate analysis revealed that children with periventricular venous infarction had lower rates of neonatal seizures (OR = 0.02, CI: 0.001-0.23), systemic thrombosis (OR = 0.09, CI: 0.01-0.96), and emergency cesarean section (OR = 0.32, CI: 0.09-0.98) (see Table 2).
Distribution of gestational age at birth. Children with perinatal arterial ischemic stroke and periventricular venous infarction were born at 37 to <43 weeks’ gestation at similar rates (P = .679). *Note 14 children were missing data on weeks of gestational age.
Multivariable Model Exploring Clinical Features of PAIS and PVI.
Abbreviations: CI, confidence interval; OR, odds ratio; PAIS, perinatal arterial ischemic stroke; PVI, periventricular venous infarction.
At or below P = .05.
Discussion
We conducted a population-based (Ontario, Canada) study exploring pregnancy and perinatal clinical features in a subgroup of children with hemiplegic cerebral palsy and either perinatal arterial ischemic stroke or periventricular venous infarction on neuroimaging. We found that neonatal systemic thrombosis, emergency cesarean section, and neonatal seizures were more commonly observed in infants with perinatal arterial ischemic stroke than with periventricular venous infarction.
Perinatal arterial ischemic stroke and periventricular venous infarction are the most common vascular brain lesions causing hemiplegic cerebral palsy. 19 Maternal health, pregnancy, obstetrical-related factors, and fetal and infant health all impact the condition of the infant brain and its vascular supply.20–23 We found more than three-quarters of infants with periventricular venous infarction were born at term, as reported by Fehlings et al. 7 A recent case-control study by Vitagliano 12 proposed the pathophysiology of periventricular venous infarction is that of an in utero germinal matrix hemorrhage and subsequent medullary vein compression occurring before 32 weeks’ gestation. Our findings indicate that if periventricular venous infarction is related to premature brain injury in hemiplegic cerebral palsy, it occurs in utero, remote from term delivery in a significant proportion of babies.
Neonates with hemiplegic cerebral palsy due to perinatal arterial ischemic stroke had higher rates of systemic thrombosis in the newborn period compared with the periventricular venous infarction group. The incidence of neonatal thrombosis has a reported incidence of 6.9 to 15 per 1000 neonatal intensive care unit admissions.24–29 The range and increase is likely due to advances in the management and survival of younger neonates and improved detection of thrombosis. In perinatal arterial ischemic stroke, vascular occlusion results from intravascular thrombosis (including embolism) intraluminal thrombotic clot, whereas in periventricular venous infarction, medullary vein occlusion is most likely related to vascular compression by mass effect from a germinal matrix hemorrhage. 30
In perinatal arterial ischemic stroke, the rate of systemic thrombosis was 7%. In these infants, thrombosis in systemic vessels may have provided the source for intracardiac paradoxical embolism into the cerebral circulation through normal fetal intracardiac shunting. Alternatively, a prothrombotic state may have predisposed infants to thrombosis in multiple locations, including cerebral arteries and systemic blood vessels. The nonsignificant trend for increased family history of thrombosis we observed in our perinatal arterial ischemic stroke group may support the contribution of an inherited prothrombotic disorder to perinatal arterial ischemic stroke. Newborn infants have an enhanced thrombotic tendency related to third-trimester maternal thrombosis, elevated fetal-newborn hematocrit, and small blood vessels, which promote thrombosis. 31
Emergency cesarean section was found to be a clinical feature associated with perinatal arterial ischemic stroke, also reported by others studying neonatal perinatal arterial ischemic stroke. 19 Emergency cesarean section is thought to represent a marker for difficulties in late gestation or labor, which themselves lead to the need for urgent assisted delivery rather than the cesarean section itself being causative for perinatal arterial ischemic stroke. 19 In a meta-analysis of cesarean delivery and cerebral palsy, emergency cesarean section in term infants was associated with an increased risk of cerebral palsy, whereas elective cesarean delivery was not. 32
We also found a higher rate of neonatal seizures in the perinatal arterial ischemic stroke compared with the periventricular venous infarction group. This corresponds with the timing of the perinatal arterial ischemic stroke event, which likely occurs acutely within days of infant delivery and likely reflects a clinical feature or symptom of acute brain injury. In other pediatric ages, acute perinatal arterial ischemic stroke is frequently associated with new-onset seizures in the first day or week and only rarely in subsequent weeks or months. This may also reflect the predominant location of perinatal arterial ischemic stroke involving the cerebral cortex, a more epileptogenic brain tissue than the deep gray or white matter tissue involved in periventricular venous infarction.
Although maternal illness was not frequently observed, we did find a nonsignificant increase in the rate of infertility treatment in periventricular venous infarction compared with perinatal arterial ischemic stroke. In vitro fertilization comprised 50% of the infertility treatment in both groups. Assisted reproductive therapy is reported to be associated with adverse perinatal outcomes in singleton infants compared to spontaneously conceived singletons. 33 Specifically, fresh embryo transfer used during in vitro fertilization treatment is associated with a higher risk of preterm birth, small for gestational age babies, 34 and double the risk of cerebral palsy. 35 Children born after assisted reproductive therapy also have altered epigenetic profiles.36,37 A significant challenge in understanding adverse perinatal outcomes after assisted reproductive therapy, including periventricular venous infarction, is differentiating the factors related to underlying biological issues leading to infertility vs the assisted reproductive therapy treatments per se. The causative pathways of infertility and assisted reproductive therapy that contribute to periventricular venous infarction are unknown to date. However, assisted reproductive therapy is also complicated by a 2-fold increased risk of maternal venous thrombo-embolism in early pregnancy, which could manifest in the transfer of prothrombotic factors to the placenta or fetus. 38 This suggests that in utero thrombosis could be a predisposing risk factor for periventricular venous infarction because of the increasing likelihood of obstruction with more minor degrees of medullary venous compression.
Our study has limitations. The brain imaging used in our analysis was acquired clinically and conducted at varying ages and intervals from brain lesion onset. Excluding children without neuroimaging (23%) could have introduced bias. Also, important variables we collected were excluded from the analysis because of more than 20% of data missing. These included family history of thrombosis, intrapartum asphyxia/hypoxic-ischemic encephalopathy, congenital malformation or syndrome, meconium, and Apgar at 5 minutes. The lack of standardized obstetrical and newborn data on these factors in routine health record charting presents a gap for retrospective research. Universal common data elements could be developed for standardized clinical history and health data recording, enabling improved completeness of data and data pooling across studies. Finally, our sample size of children with hemiplegic cerebral palsy was moderately sized, with imaging patterns of periventricular venous infarction in 49 and perinatal arterial ischemic stroke in 95, limiting the power and the number of variables for multivariate analysis. Despite these limitations, our strengths include the provincewide representation of hemiplegic cerebral palsy patients, data collection from multiple sources (chart review, parent report, source imaging studies), and the implementation of a neuroimaging classification and scoring system for periventricular venous infarction and perinatal arterial ischemic stroke conducted by pediatric neuroradiologists and having excellent interrater reliability.
Clinical implications from our findings include the previously noted recognition that periventricular venous infarction is frequently present in term-born infants 7 and that increased diagnostic suspicion for perinatal arterial ischemic stroke should be considered in newborns diagnosed with systemic thrombosis. Although periventricular venous infarction is well visualized on cranial ultrasonography, cross-sectional imaging with MRI is needed to diagnose perinatal arterial ischemic stroke confidently. Given the association of systemic thrombosis and perinatal arterial ischemic stroke, it may be reasonable to consider an MRI in newborns diagnosed with systemic thrombosis to detect clinically silent perinatal arterial ischemic stroke. No studies have been done to determine the incidence of clinically silent perinatal arterial ischemic stroke in newborns with systemic thrombosis; however, we found systemic thrombosis in 7% of perinatal arterial ischemic stroke patients. In addition, attention to fetal-maternal health focused on prothrombotic factors in the days leading up to delivery could lead to strategies to reduce the frequency of neonatal perinatal arterial ischemic stroke.
Our study examined clinical features associated with perinatal arterial ischemic stroke or periventricular venous infarction in a geographically defined population, including a spectrum of preterm- to term-born children with hemiplegic cerebral palsy. Our study suggests that acute obstetrical factors and concomitant systemic thrombosis are associated with the perinatal arterial ischemic stroke pattern in children with hemiplegic cerebral palsy. Our finding that clinical features in children with hemiplegic cerebral palsy differ depending on the type of brain injury, whether periventricular venous infarction or perinatal arterial ischemic stroke, when taken together with recent population-based studies,2,12 emphasizes the divergent causal pathways that ultimately lead to hemiplegic cerebral palsy. Future strategies to prevent hemiplegic cerebral palsy will benefit from further elaborating on these distinct causal pathways.
Footnotes
Acknowledgments
The authors wish to thank the participants and families who donated their time to this research study and the numerous research staff, occupational therapists, psychologists, and physicians across Ontario, Canada, who contributed their time, clinical expertise, and support.
Author Contributions
D.F. conceived of the original idea, study idea and design, in charge of overall direction, planning, and funding acquisition for the study. GdeV., contributed to the design and implementation of the manuscript, to the analysis of the results and to the writing of the manuscript. T.D. was responsible for data analysis and interpretation and the lead writer of the manuscript. A.R. was responsible for logistics of data transfer and organization of data for analysis. P.K, M.S., M.M, D.F. and GdeV. developed the neuroimaging classification form. P.K. and M.M. validated and scored the neuroimaging classification form. N.D. and all authors, including members of the CP-NET group contributed data, discussed the results, and contributed comments and edits for the final version of the manuscript.
Author Note
The members of the CP-NET group are as follows: Craig Campbell (Department of Paediatrics, Western University, London, ON, Canada); Anne Kawamura and Lauren Switzer (Department of Paediatrics, Bloorview Research Institute, Holland Bloorview Kids Rehabilitation Hospital, University of Toronto, Toronto, ON, Canada); Dawa Samdup (Department of Pediatrics, Queens University, Kingston, ON, Canada); and Ilana Walters (Faculty of Medicine, Dalhousie University, Halifax, NS, Canada), Carolyn Hunt (Department of Paediatrics, Grandview Children's Centre, University of Toronto, Toronto, ON, Canada) Marie Kim (Erin Oak Kids Center for Treatment and Development, Mississauga, ON, Canada), Ronit Mesterman (Department of Pediatrics, CanChild, McMaster Children's Hospital, McMaster University, Hamilton, ON, Canada), Jan Willem Gorter (Department of Pediatrics, CanChild, McMaster Children's Hospital, McMaster University, Hamilton, ON, Canada), Anna McCormick (Department of Pediatrics, the Children's Hospital of Eastern Ontario [CHEO], Ottawa, ON, Canada).
Data Availability Statement
The data supporting this study's findings are available from the corresponding author on reasonable request.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was conducted with the support of the Ontario Brain Institute, an independent non-profit corporation funded partially by the Ontario government. The opinions, results, and conclusions are those of the authors and no endorsement by the Ontario Brain Institute is intended or should be inferred. Furthermore, the Ontario Brain Institute was not involved in the study design, data collection, data analysis, manuscript preparation, or publication decisions.
