Population Based Birth Prevalence of Disease-Specific Perinatal Stroke

Cases were identified by retrospective and prospective methods within the Alberta Perinatal Stroke Project.¹⁴  This population-based research cohort was designed to capture all perinatal stroke cases in southern Alberta, Canada (population ∼2.1 million) over a 28-year period (1990–2017). The Alberta Perinatal Stroke Project leveraged Canadian universal health care and a single tertiary care pediatric center (Alberta Children’s Hospital; ACH) where all children with neurologic presentations or disabilities within the southern provincial population were assessed. All cases were documented in the program registry for diagnosis and date of birth. Methods were approved by the Conjoint Research Ethics Board at the University of Calgary.

Cases were term-born (>36 weeks) children born with consistent clinical history and neuroimaging-confirmed perinatal stroke. Prospective hot pursuit¹⁵  involved identifying patients while under care and was used to identify acute cases (NAIS, CSVT, NHS) during their hospitalization at ACH or other centers prospectively from 2008 to 2017 through referral to an established pediatric stroke consult service by both clinical and radiographic presentations (Supplemental Fig 6). All 22 perinatal centers in southern Alberta were included, with the majority of referrals from the 4 tertiary NICUs in Calgary. As per protocol, the perinatal care providers for symptomatic infants contacted 1 of the 4 tertiary or quaternary NICUs or the ACH pediatric neurology service, who then guided either further investigations or transport to a tertiary center. Once stroke was identified, referral was made to 1 of the 2 pediatric stroke neurologists. Prospective outpatient cases (APPIS, PVI, PPHS) were identified through referral to the ACH pediatric stroke clinic by family physicians, rehabilitation specialists, pediatricians, 2 community pediatric neurologists, and the 5 to 10 pediatric neurologists at ACH, as well as other pediatric specialists. Diagnostic MRI was obtained per a standardized institutional stroke protocol, and the stroke diagnosis was confirmed and classified after review of the images by a pediatric neuroradiologist and pediatric stroke specialist according to the criteria below. Prospective cases were recorded in a secure database by the 2 pediatric stroke neurologists (AK and AM) after being consulted in hospital or in clinic.

Cold pursuit cases of NAIS, APPIS, and PVI were retrospectively identified from 1990 to 2012 by strategic searching of 78 different ICD-9 and ICD-10 codes (Supplemental Table 3). A medical data specialist was contracted to perform unbiased, systematic searches of all possible data sources within the population, for both inpatients and outpatients. All matches were cross-referenced for duplicates, after which a complete list of possible cases was delivered to the research team. Five research students underwent detailed, validated training on how to identify potential perinatal stroke cases from medical chart information focusing on essential clinical elements (eg, term birth, neonatal neurologic presentations, childhood presentations of cerebral palsy) and radiology report details. All information was scored on a standardized data capture form. After this, students would meet in person with the study coordinator and principal investigator to present their cases, either identifying a clear exclusion criterion or outlining reasons the case was a potential perinatal stroke. The coordinator then sought to obtain original neuroimaging for all potential cases. This included accessing online digital images (and any child reimaged since 2000) or ordering the original hardcopy films for all patients with older imaging. Once obtained, the imaging studies were then reviewed by the principal investigator for diagnostic conformation and classification.

There was no retrospective search for cases of CSVT. Cases of NHS and PPHS were retrospectively identified from 1992 to 2008 using similar methods above, and published by Cole et al.¹⁶  Only cases that were confirmed with imaging and evaluated with a history and physical examination in our specialty clinic were ultimately included for birth-prevalence calculation. Cases were categorized according to the imaging criteria below.

Diagnostic neuroimaging review confirmed perinatal stroke and classified the specific subtype on the basis of established imaging criteria.^7,14,17  The following disease states were defined:

• 1. NAIS. One or more areas of restricted diffusion within an arterial territory.

• 2. CSVT. Presence of thrombus within a cerebral vein or sinus confirmed by MRV or CTV or, if a thrombus cannot be definitively demonstrated, a pattern of venous infarction within a known cerebral venous territory.

• 3. NHS. Focal collection of acute or subacute blood within the brain parenchyma with or without intraventricular or subarachnoid blood. Cases in which blood was exclusively extra-axial were not included. Hemorrhagic transformation of ischemic strokes were classified as their primary disease.

• 4. APPIS. Chronic, focal encephalomalacia attributable to remote infarction in an arterial territory.

• 5. PVI. Chronic, focal encephalomalacia of the periventricular white matter with sparing of the deep gray matter and cortex. Additional supportive evidence included presence of remote germinal matrix hemorrhage demonstrated by areas of hypointensity on susceptibility weighted imaging.

• 6. PPHS. Evidence of remote intraparenchymal hemorrhage on imaging.

Additional variables collected were infant sex, stroke side, affected vessel, and age at diagnosis.

Both live and still births were included in the birth-prevalence calculation. The number of live births in southern Alberta was determined by using the Alberta census¹⁸  for the years 1996–2017. For the years 1990–1995, live births for southern Alberta were extrapolated from the Alberta live births by using the average proportion of live births that occur in Southern Alberta (0.481). The number of still births was calculated by using the provincial registry. The proportion of term births in Alberta were used for each year for 2000–2017,¹⁹  and for 1990–2000, the average proportion (0.91) was used to generate the total term births for each year.

The total estimated birth prevalence of perinatal stroke of all subtypes was calculated by summing all cases for 2008–2017 over all births for 2008–2017 because of the consistent method of case acquisition in that period.

Before estimating birth prevalence for each stroke subtype, changes over time were assessed by calculating annual birth prevalence and using Poisson regression in cases of at least 20 data points. Assumptions were tested by assessing for overdispersion and assessment of the residuals to ensure Poisson distribution. Wilcoxon rank-sum test was performed to determine if the annual birth prevalence before and after 2008 was from populations with the same distribution. Medians were expressed with interquartile ranges (IQRs; 75th centile to 25th centile). When the estimated annual birth prevalence for the perinatal stroke subtype was stable over time, the estimated birth prevalence was determined by summing all cases and dividing by all births during the same period. When there were significant changes over time, the 2008–2017 prospective data were used. Birth prevalence was also estimated for each sex. The sensitivity of each ICD code was determined by dividing the confirmed cases (true-positives) by all the cases identified by that code.

The age at imaging diagnosis for late-presenting perinatal stroke diseases (APPIS and PVI) was evaluated by linear regression if assumptions were met. Wilcoxon rank-sum test was performed to determine if the annual birth prevalence before and after 2008 were from populations with equal distributions.

Binominal test of proportions was used to determine if there were significant differences in perinatal stroke birth prevalence between the sexes, with the expected proportion of boys compared with girls at 0.51, based on the birth ratios in Alberta from 2000 to 2017 (mean: 0.513, SD: 0.029, range: 0.505–0.520).²⁰  In addition, a two-sample proportions test was used to compare sex-specific incidence of perinatal stroke. Comparison of NAIS and APPIS by sex was performed with Fisher exact test. Test of proportions was used to compare the proportion of left-sided strokes to total of left and right, with expected proportion of 0.5, and α was set at 0.05. Analysis was by Stata (StataCorp, College Station, TX) version 15.1.

A total of 359 perinatal strokes were identified. Of these, 293 (82%) were diagnosed after prospective case acquisition started in 2008, whereas 66 (18%) were diagnosed before 2008 (Fig 1A). The median number of perinatal strokes diagnosed annually before 2008 was 2 (IQR = 3), and the median after 2008 was 28 (IQR = 9), P < .0001 (Fig 1B). Perinatal stroke subtype birth prevalence and time periods are summarized in Tables 1 and 2 and described below.

Total term-born perinatal stroke birth prevalence was calculated for birth years 2008–2017. There were 219 cases of perinatal stroke identified, consisting of 80 NAIS (37%), 47 PVI (21%), 35 NHS (16%), 26 APPIS (12%), 26 CSVT (12%), and 5 PPHS (2%) (Fig 2A). From 236 832 term births during the same time period, the total cumulative estimated birth prevalence of perinatal stroke was 1:1100 (95% confidence interval [CI]: 1:1300–1:900) (Table 1). Acute perinatal stroke presentations totaled 141 of 219 (64%), and delayed presentations totaled 78 of 219 (36%) (Fig 2A). Arterial causes represented 146 of 219 (67%), whereas venous represented 73 of 219 (33%) (Fig 2B). The median annual birth prevalence was significantly higher after 2008 for NAIS (P < .0001) and NHS (P = .025) (Fig 3).

There were 107 cases of NAIS in 1990–2017, with significantly more cases identified after 2008 (P < .0001, Fig 3A). Therefore, the estimated birth prevalence was calculated by using the 80 cases from 2008 to 2017 and was 1:3000 (95% CI: 1:3800–1:2400). NAIS represented 37% of all perinatal stroke, 55% of arterial stroke, and 57% of acutely presenting perinatal stroke (Fig 2). NAIS were left-sided in 59 cases (55%), right-sided in 26 (24%), with a left/right ratio of 0.69 (P = .0003), and bilateral in 22 (21%) (Table 2). In 83 cases, a single arterial distribution was affected (78%), whereas 17 had 2 arteries affected (16%), and 7 had ≥3 (7%). The middle cerebral artery (MCA) was the most frequently affected distribution (95% of single artery infarcts, 82% of all infarct distributions), with single artery distributions of 68% left MCA and 32% right MCA (P = .0015). The posterior cerebral (16 of 116, 14%) and basilar artery (2 of 116, 2%) were less commonly affected.

There were 26 cases of neonatal CSVT from 2008 to 2017, for an estimated birth prevalence of 1:9100 (95% CI 1:16 000–1:6400, Fig 3B, Table 1). CSVT represented 12% of perinatal stroke, 36% of venous stroke, and 18% of acutely presenting perinatal stroke (Fig 2).

There were 54 cases of NHS identified in 1992–2017, and annual birth prevalence over time was stable (P = .18); however, the median annual birth prevalence was significantly higher after 2008 compared with before 2008 (1:6800 vs 1:15 000, P = .025). Therefore, the 35 cases from 2008–2017 were used to estimate the birth prevalence of 1:6800 (95% CI: 1:14 000–1:4400) (Fig 3C, Table 1). NHS represented 16% of perinatal stroke, 24% of arterial stroke, and 25% of acutely presenting stroke (Fig 2).

There were 71 cases of APPIS in 1990–2017, with stable annual birth prevalence over time (P = .45, Fig 3D). The estimated birth prevalence of APPIS was 1:7900 (95% CI: 1:11 000–1:6200). APPIS represented 12% of perinatal stroke, 18% of arterial stroke, and 33% of delayed presenting perinatal stroke (Fig 2). A single arterial territory was affected in 65 of 71 (92%) patients. The left side was affected in 40 (56%) and the right in 27 (38%, P = .11) cases, and 4 cases were bilateral (6%). In cases of a single affected artery, the MCA was involved in 62 (95%), with the left affected in 38 (61%) and the right in 24 (39%, P = .075) (Table 2). There were 3 cases affecting the PCA (5%), 2 of bilateral MCA involvement, and 2 of multifocal stroke.

There were 93 cases of PVI in 1990–2017, with a stable annual birth prevalence over time (P = .66, Fig 3E) The estimated birth prevalence of PVI was 1:6000 (95% CI 1:8000–1:4800). PVI represented 22% of perinatal stroke, 64% of venous stroke, and 60% of delayed presentation stroke (Fig 2). PVI occurred on the left in 54 (58%), on the right in 38 (41%, P = .095), and was bilateral in 1 (1%) (Table 2).

There were 8 cases of PPHS in 1992–2017, for an estimated birth prevalence of 1:65 000 (95% CI 1:234 000–1:38 000) (Fig 3F). PPHS represented 2% of perinatal stroke, 3% of arterial stroke, and 6% of delayed presenting perinatal stroke (Fig 2).

The birth prevalence of perinatal stroke in term males was 1:860 (95% CI: 1:960–1:790), which was significantly higher than in females (1:1500, 95% CI: 1:1800–1:1200, P = .0001) (Supplemental Table 4). Of those with perinatal stroke, the proportion of males was 225 of 359 (63%, 95% CI: 58%–68%, P < .0001, Table 2). The proportion of males was significantly higher for NAIS, NHS, and PVI, and there was a trend toward more males affected for CSVT (Fig 4). The birth prevalence of acute presentations of neonatal stroke was 1:1300 for males and 1:2600 for females (P < .0001), whereas for delayed presentation, the birth prevalence was 1:2800 for males and 1:3400 for females (P = .37). Males with arterial ischemic stroke presented acutely with NAIS twice as often as delayed presentation with APPIS (67% vs 33%), whereas for females, the proportions were evenly divided between acute presentation (49%) and delayed (51%) (P = .027).

There were 164 children diagnosed with APPIS and PVI, 50% born before 2008 and 50% after, although 17 (10%) were diagnosed before 2008, whereas 147 were diagnosed after (90%) (Fig 5). The age at diagnosis for PVI and APPIS did not meet the assumptions for linear regression. The median age at diagnosis of APPIS was 10.6 years for children born before 2008 compared with 1.4 years for children born after 2008 (P < .0001, Fig 5C). The median age at diagnosis of PVI was 8.5 for children born before 2008 compared with 1.4 years for children born after 2008 (P < .0001, Fig 5D).

A total of 76 ICD-9 and ICD-10 codes were investigated for NAIS, APPIS, and PVI from 1990 to 2012, yielding 1790 results evaluated for perinatal stroke (Supplemental Table 3). There were 180 of 1790 (10%) cases of perinatal stroke of the following subtypes: PVI, 57 (32%); NAIS, 38 (21%); NHS, 35 (19%); APPIS, 33 (18%); and CSVT, 12 (7%). The highest yield codes had sensitivities of 44% to 53%. Codes that would appear sensitive, such as “cerebral arterial occlusion not otherwise specified with infarct,” only yielded a diagnosis of stroke in 7 of 20 (35%) of cases.

The overall estimated term-born birth prevalence of 1 case per 1100 was higher than previous estimates and even higher for male infants, at 1 case per 860. Previous estimates are between 1:1600 and 1:8000.^8–10  Our estimate may be higher because of higher sensitivity of prospective methods,²¹  more specific and inclusive classifications of perinatal stroke,⁶  and the improved ability to accurately diagnose perinatal stroke afforded by sensitive MRI neuroimaging modalities, including diffusion for acute stroke, venography for CSVT, and susceptibility weighted images for PVI.^7,22  We are careful to note that our results still reflect the minimum estimates of birth prevalence in term-born children.

The estimates of birth prevalence in individual stroke subtypes are comparable to previous studies. The estimated birth prevalence of NAIS was higher than previous estimates of 1:20 000 to 1:5600,²³  whereas the estimated birth prevalence of PVI was slightly lower than one previous estimate of 1:3400.²⁴  The birth prevalence of APPIS was similar to previously reported 1:7100,²⁴  and CSVT was also in the range of previously reported birth prevalence in neonates of 1:50 000 to 1:8300.^25,26  Both NHS and PPHS birth prevalence have been reported by our group.¹⁴ 

The relative proportions of perinatal stroke syndromes we observed are not well reflected in the overall literature. For example, as a disease, NAIS dominates the literature in perinatal stroke, likely accounting for >80% of published cases; however, <40% of our cases were NAIS. In contrast, the presumed perinatal strokes of APPIS and PVI constitute a large proportion of cases, yet there are fewer articles in the literature,^7,27  and there are no controlled risk factor studies. These relative imbalances suggest potential gaps in research efforts in which additional attention on understudied specific diseases may be indicated.

An important finding in our study was the significantly higher estimated birth prevalence of NAIS and NHS coincident with changing from retrospective to prospective case acquisition. Other reasons for increased birth prevalence include the increased availability and diagnostic accuracy of MRI for symptomatic neonates over the study period, as well as an increased awareness of stroke as a cause of acute neurologic symptoms in neonates. Although it is possible the true birth prevalence increased over the time period, these other important factors preclude this conclusion. The disparity between prospective and retrospective case identification in cerebrovascular disease has been previously shown,²¹  and our study reinforces the potential pitfalls of using purely retrospective measures to generate estimates of birth prevalence.

We confirm previous descriptions of a male predominance in most forms of perinatal stroke. In NAIS, PVI, and NHS, the proportion of males was higher, with a similar trend for CSVT. This is consistent with multiple previous studies of more varied populations demonstrating a male preponderance.^28–30  We were surprised to find that in arterial ischemic stroke (NAIS, APPIS), males were more likely to present acutely than females. This may be due to greater vulnerability of male neonates to NAIS risk factors, such as meningitis,³¹  or difficult transition to postnatal life,¹⁴  such as fetal bradycardia, emergency caesarian, and low Apgar scores,^32,33  or possibly a lower seizure threshold that makes acute detection more likely. In addition, the greater proportion of male infants affected by PVI, a fetal stroke, is seen in germinal matrix bleeding in delivered premature infants.³⁴  Sex-specific differences in perinatal stroke require careful consideration in future studies.

We demonstrate a marked reduction in delay to diagnosis of presumed perinatal stroke after institution of an organized pediatric stroke program. The explanation for this is probably multifaceted, including an evolving emphasis on obtaining diagnosis in children with cerebral palsy by imaging³⁵  and increased referrals of children who may have been previously under-investigated. The data did not meet the assumptions for linear regression; however, the approximately linear relationship is partially due to >80% of the diagnoses occurring after 2008 because age at diagnosis is calculated as diagnosis year – birth year. Prompt diagnosis is translationally relevant,^36,37  and rationale includes earlier access to emerging developmental therapies but also alleviation of potential long-term mental health morbidity in parents and families.^5,38 

Important limitations include how cases were identified. Selection bias for more severe (and more clinically relevant) cases probably existed. We intentionally did not address premature infants, although strokes certainly can occur in that age group.^39,40  In particular, germinal matrix hemorrhage–intraventricular hemorrhage with periventricular hemorrhagic infarction is an important cause of stroke in preterm infants. Our estimate of PVI birth prevalence does not include these infants, and thus only reflects in utero periventricular hemorrhagic infarction occurring in term-born infants. Prospective case identification of cases presenting in peripheral southern Alberta centers was dependent on referral to our center; however, there are established pathways for referrals to our hospital (Supplemental Fig 6). There were no delayed cases identified that on retrospection had been diagnosed as symptomatic neonates but not referred to the stroke program, supporting that the majority of acute symptomatic cases were captured by the referral system. For CSVT, diagnosis requires the use of venous imaging, which has not always been routinely performed. The retrospective case acquisition is limited by the correct coding of the diagnosis for ICD-9 and ICD-10 codes and our review methods.

The overall birth prevalence of perinatal stroke in term-born children is 1:1100 but should always be considered as one of many specific disease states. Organized stroke programs and modern neuroimaging improve the accuracy and promptness of perinatal stroke diagnosis while administrative data sets are not reliable. Because perinatal stroke is a leading cause of lifelong disability and much is still unknown regarding pathophysiology, prevention, or treatment, disease-specific perinatal stroke research is urgently required.

ACH

Alberta Children’s Hospital

APPIS

arterial presumed perinatal ischemic stroke

CI

confidence interval

CSVT

cerebral sinovenous thrombosis

ICD

International Classification of Diseases

IQR

interquartile range

MCA

middle cerebral artery

NAIS

neonatal arterial ischemic stroke

NHS

neonatal hemorrhagic stroke

PPHS

presumed perinatal hemorrhagic stroke

PVI

periventricular venous infarction