Cerebral palsy (CP) is one of the most common causes of childhood physical disability.¹ Unilateral CP is the most frequent subtype in infants born at term. Most cases of unilateral CP result from perinatal stroke² and such disease specificity is required to explore the pathophysiology and advance prevention.

Perinatal strokes are specific, vascular brain injuries occurring in the fetus or neonate. Acute symptomatic varieties typically present with seizures in the first days after birth and include neonatal arterial ischemic stroke (NAIS), neonatal hemorrhagic stroke, and focal lesions associated with cerebral sinovenous thrombosis, all of which account for approximately 50% of perinatal strokes.³ Presumed perinatal strokes, such as periventricular venous infarction (PVI) and arterial presumed perinatal ischemic stroke (APPIS), present later in infancy or childhood, usually with unilateral CP. The most common varieties are arterial ischemic stroke (AIS) and PVI.⁴

The causative mechanisms of perinatal stroke remain poorly understood. AISs are the most thoroughly studied and many are secondary to placental thromboembolism.^5-7 Case–controlled studies have explored possible AIS risk factors, but associations have been inconsistent.^{4, 5, 8-13} The most consistent relate to perinatal adversity including emergency Caesarean section,^{4, 9-12} resuscitation,^{4, 10-12} fetal heart rate abnormality,^{4, 9, 11, 12} meconium staining,^{4, 8, 9, 11, 12} prolonged second stage of labor,^{4, 9, 11, 13} and Apgar score of less than 7 at 5 minutes.^{4, 9-12} However, such general associations fail to define clear mechanisms. PVIs are thought to occur in utero before 32 weeks’ gestation, secondary to germinal matrix hemorrhage (GMH) and secondary venous infarction of the periventricular white matter.^{14, 15} Controlled risk factor studies have not been completed for PVI.¹⁵

There is a pressing need to explore disease-specific risk factors for perinatal stroke. Without them, options to prevent perinatal stroke remain limited. We employed a large, controlled sample to examine clinical risk factor profiles associated with AIS and PVI. We hypothesized that profiles would differ between groups, with intrapartum factors being associated with AIS.

METHOD

RESULTS

From the period February 2003 to January 2019, the CCPR enrolled a total of 2093 participants of whom 662 (32%) were classified as having unilateral CP. From this, 152 cases (23%) were excluded because of lack of an MRI report. No apparent variation in case type or severity of perinatal stroke was observed across the country. Approximately 15% of original MRIs were reviewed between two investigative sites, where classification of stroke and stroke type demonstrated 100% agreement with the original classification based on the radiology report alone.

Of the remaining 510 cases, 342 (67%) had isolated AIS or PVI. Of these, 335 (98%) had sufficient data (>90% fields complete out of 125 variables) available for analyses, whereby 36 cases were excluded as they were born before 35 weeks’ gestation. The remaining 299 cases constituted the primary population for analysis. AIS was diagnosed in 169 (57%) and PVI in 130 (43%) of the eligible cases. Probable NAIS occurred in 36 of 169 (21.3%) cases and probable APPIS in 128 of 169 (78.7%). Median age at diagnosis for AIS was 11.9 months (interquartile range: 6.2–25.7mo; range 0.17–104.1mo), and 58.6% were male. Median age at diagnosis for PVI was 25.3 months (interquartile range: 14.5–38mo, range 0.7–114.7mo) and 57.7% were male. Both arterial and venous populations had a male predominance, though neither reached significance: 99 of 169 (59%, p=0.13) and 75 of 130 (60%, p=0.25) respectively.

DISCUSSION

The goal of this study was to explore the clinical factors associated with disease-specific perinatal stroke. Our primary findings suggest that stroke subtypes carry unique risk factor profiles, consistent with their different appearance, timing, and hypothesized pathophysiology. Consistent with previous investigations,^{4, 8-12} we found that AIS was associated with intrapartum variables, supporting mechanisms surrounding the time of birth. As the first controlled study of factors associated with PVI, a profile consistent with timing earlier in gestation was observed.^{14, 15} Our results demonstrate the potential of large, controlled registries to investigate disease-specific mechanisms of CP. Our observed association between non-specific, birth-related factors and AIS was expected.

Multiple controlled studies of NAIS,^{4, 8-12} ranging in sample from 12 to 134 cases, have suggested similar results, though specific factors have been inconsistent. While certainly supporting the fact that many AIS lesions occur close to birth, it is erroneous to suggest that ‘hypoxia’ is a primary causative mechanism, as suggested in previous meta-analyses.^{22, 23} An alternative hypothesis for why such factors might be associated with both hypoxic–ischemic encephalopathy and perinatal AIS is a shared risk imparted to the fetus for a difficult transition, such as placental disease.

A leading hypothesis for the pathogenesis of AIS is that of placental thromboembolism. Placental inflammation may be a driver of this risk.^{5, 24, 25} Placental conditions such as chorioamnionitis, fetal thrombotic vasculopathy, and villitis of unknown origin may lead to thromboembolization into the fetal circulation.^{4, 26} Our findings suggesting a relationship between chorioamnionitis and AIS support this proposed mechanism. Unfortunately, difficulty in obtaining placental tissue after birth has limited the pathological studies that could demonstrate a clear association with AIS.²⁷ The way in which chorioamnionitis is captured as a variable in the CCPR does not distinguish whether the diagnosis was based on pathological tissue or simply a clinical diagnosis. This limits the extent to which clear conclusions can be drawn with regards to placental thromboembolism and AIS. Somewhat surprisingly, PROM demonstrated a potentially protective effect for AIS (5.8% vs 12.3% of controls). One possible explanation is that mothers with PROM may have received a prophylaxis dose of antibiotics, which may have been protective in this inflammatory disease state.

Our grouping of both acute and remote presenting AIS is both a strength and weakness. Risk factors for APPIS were essentially unknown with no controlled study to date. One small descriptive study suggested the possible occurrence of factors previously associated with perinatal stroke, but modest power and absence of controls prevented any speculation about the truth or strength of the correlation.¹⁵ Arterial ischemia is the most common mechanism of perinatal stroke, approximately 75% of which probably presents acutely as NAIS.²⁸ Of those children, only a minority will develop CP.²⁹ The remaining 25% present later in childhood, usually with motor impairment and are diagnosed with CP. Thus, we would expect that, of children with CP and AIS, approximately 70% would have APPIS and 30% NAIS (Fig. S1, online supporting information). Indeed, by using the CCPR and selecting only those children with atypical motor outcomes, the majority of cases with AIS were also APPIS (128/164, 78%), providing a unique opportunity to include this group that is often neglected in AIS risk factor analysis. That the chronic images of both acute and late presenting arterial strokes are virtually indistinguishable suggests they are likely to reflect the same disease. How our inclusion of late-presenting AIS influences the associations we observed cannot be accurately determined but may explain some differences from previous studies of NAIS only. For example, primigravida was not observed as a risk factor for AIS in our study, but has been reported in multiple previous studies, though inconsistently.⁴ Perhaps primigravida predisposes females to present with intrapartum complications, which themselves may correspond with AIS. Ongoing studies of APPIS should shed light on this issue.

This was the first controlled, risk factor analysis of PVI. Congruent with evidence that venous strokes occur earlier in gestation,^{14, 15} we found associations with a predominance of in utero factors. The proposed pathophysiology of PVI is that of an in utero GMH and subsequent medullary vein compression occurring before 32 weeks’ gestation. Unlike GMH in infants born preterm, where peripartum and intensive care stressors are known risk factors, it is unclear why GMH would occur in the fetus. Our association between infants who are SGA and have PVI is non-specific but potentially plausible, perhaps suggesting co-occurrence of chronic stressors on the fetus. A recent large meta-analysis of antenatally-diagnosed GMH by our group demonstrated an uncomplicated pregnancy in only 38% of PVI cases and several potential co-occurring conditions, including twin gestation and congenital anomalies, but also intrauterine growth restriction.³⁰ We found primigravida was protective in PVI, which may be relevant if a subset of PVI is related to conditions such as fetal alloimmune thrombocytopenia, which has been associated with intracranial hemorrhage in multigravida pregnancies due to alloimmunization.³¹

Unexpectedly, the PVI group did show some associations with intrapartum variables, specifically emergency Caesarean section and chorioamnionitis. This could be due, in part, to our limited definition in controls of any Caesarean section that was not elective, indicating that the final numbers may be an overestimation. Alternatively, it may reflect common mechanisms that result in PVI and peripartum fetal distress, such as placental pathology. Chorioamnionitis as a risk factor for PVI could be due to multiple mechanisms. Chorioamnionitis occurring near the end of a late preterm delivery at 35 or 36 weeks may occur before the involution of the germinal matrix and result in the quintessential PVI. Supporting this, we found increased odds of PVI in infants born late preterm compared to at term. Another plausible explanation for chorioamnionitis increasing the odds of PVI is the possibility that a subset of these PVI cases were infants who sustained their injury in the peripartum period due to cerebral sinus vein thrombosis of the medullary veins, for which chorioamnionitis is a risk factor.³² There is increasing evidence to suggest that the initial detection of such strokes is under-recognized and later falsely classified as fetal PVI.³³ Future risk factor studies highlighting unique profiles associated with fetal PVI versus neonatal cerebral sinus vein thrombosis are necessary to shed further light on such misdiagnoses.

Our observed association between maternal substance use and both AIS and PVI merits specific consideration. First is the consideration of a false positive result. The reported substance use in the control group was slightly lower (0.7%) than the 1% to 5% reported average in pregnant Canadians.³⁴ Perhaps our control group was more health-conscious, considering Alberta’s high socioeconomic status and the selection bias toward more educated mothers. However, the rate of maternal substance use was also higher than the Canadian average amongst both stroke types, and the proportion of pregnancies with alcohol use was the same, arguing against an or abstaining control group. Substance abuse has not been systematically investigated in perinatal stroke. The most studied illicit drug has been cocaine, though evidence on outcomes is of very low quality and reports are conflicting.³⁵ There are many hypothesized theories which include suggestions that certain drugs may be damaging to fetal vessels or placental perfusion.³⁵ Unfortunately, agent-specific data were not available in our sample and would be important in future studies.

Translational implications of our findings include the counselling of parents. Receiving a diagnosis of perinatal stroke has serious consequences on parental mental health, including increased rates of depression and post-traumatic stress disorder.³⁶ Given that the timing of the stroke may be in the peripartum period, mothers may falsely believe they may have done something wrong during their pregnancy, leading to misplaced feelings of blame and guilt. The importance of education and support from health care providers for families of infants with perinatal stroke, therefore, becomes crucial and our results may add valuable content for this discussion. The possible roles of maternal smoking and substance use should obviously be carefully considered, and we would certainly advocate for better evidence before suggesting to a mother that such exposures may have contributed to the child’s stroke.

Our large, nationwide, controlled sample is both an asset and a liability. Our variables held inconsistent levels of reliability and were subject to selection and self-reporting biases. Future studies with greater representation across sociodemographic variables for controls, or larger prospective cohort studies, would enhance the generalizability of results. Given the complexity of our proposed models of interrelated clinical factors, interpreting associations is challenging. Failing to adjust for intermediate variables quickly biases the associations found, and this becomes especially relevant in perinatal epidemiology.²⁰ It is possible that such limitations have contributed to wider confidence intervals for some of our variables, further highlighting the discretion necessary in interpreting the results. Given the exploratory nature of our study, being transparent about such intricacies is paramount in guiding more definitive studies.

Our results suggest there are disease-specific differences in clinical factors associated with perinatal stroke. We validated previous associations between AIS and intrapartum factors with the addition of potentially modifiable maternal substance use. We have provided the first controlled data suggesting possible fetal factors in PVI. Causal models in perinatal stroke remain complex and multifactorial and true causation is rarely self-evident.