CT indicates computed tomography; MRI, magnetic resonance imaging. *Patients
with a neuroimaging report containing a keyword suggestive of stroke, or who
carried a physician diagnosis of stroke, cerebral palsy, neonatal seizures,
or a related diagnosis.
Lee J, Croen LA, Backstrand KH, Yoshida CK, Henning LH, Lindan C, Ferriero DM, Fullerton HJ, Barkovich AJ, Wu YW. Maternal and Infant Characteristics Associated With Perinatal Arterial Stroke in the Infant. JAMA. 2005;293(6):723-729. doi:10.1001/jama.293.6.723
Author Affiliations: Departments of Neurology
(Ms Lee, Mssrs Backstrand and Henning, and Drs Ferriero, Fullerton, Barkovich,
and Wu), Pediatrics (Drs Ferriero, Fullerton, Barkovich, and Wu), and Radiology
(Dr Barkovich), University of California, San Francisco; and Kaiser Permanente
Division of Research (Dr Croen and Ms Yoshida) and Department of Radiology
(Dr Lindan), Oakland, Calif.
Context Perinatal arterial ischemic stroke (PAS) is a common cause of hemiplegic
cerebral palsy. Risk factors for this condition have not been clearly defined.
Objective To determine maternal and infant characteristics associated with PAS.
Design, Setting, and Patients Case-control study nested within the cohort of all 199 176 infants
born from 1997 through 2002 in the Kaiser Permanente Medical Care Program,
a managed care organization providing care for more than 3 million residents
of northern California. Case patients were confirmed by review of brain imaging
and medical records (n = 40). Three controls per case were randomly
selected from the study population.
Main Outcome Measure Association of maternal and infant complications with risk of PAS.
Results The population prevalence of PAS was 20 per 100 000 live births.
The majority (85%) of infants with PAS were delivered at term. The following
prepartum and intrapartum factors were more common among case than control
infants: primiparity (73% vs 44%, P = .002),
fetal heart rate abnormality (46% vs 14%, P<.001),
emergency cesarean delivery (35% vs 13%, P = .002),
chorioamnionitis (27% vs 11%, P = .03),
prolonged rupture of membranes (26% vs 7%, P = .002),
prolonged second stage of labor (25% vs 4%, P<.001),
vacuum extraction (24% vs 11%, P = .04),
cord abnormality (22% vs 6%, P = .01),
preeclampsia (19% vs 5%, P = .01), and
oligohydramnios (14% vs 3%, P = .01). Risk
factors independently associated with PAS on multivariate analysis were history
of infertility (odds ratio [OR], 7.5; 95% confidence interval [CI], 1.3-45.0),
preeclampsia (OR, 5.3; 95% CI, 1.3-22.0), prolonged rupture of membranes (OR,
3.8; 95% CI, 1.1-12.8), and chorioamnionitis (OR, 3.4; 95% CI, 1.1-10.5).
The rate of PAS increased dramatically when multiple risk factors were present.
Conclusions Perinatal arterial ischemic stroke in infants is associated with several
independent maternal risk factors. How these complications, along with their
potential effects on the placenta and fetus, may play a role in causing perinatal
stroke deserves further study.
Perinatal arterial ischemic stroke (PAS) is an important cause of cerebral
palsy and other neurologic disabilities, including epilepsy and cognitive
impairment.1- 7 Arterial
ischemic stroke is diagnosed primarily in neonates born at term8,9 and
is responsible for 50% to 70% of congenital hemiplegic cerebral palsy in this
The cause of PAS is poorly understood. Investigators have reported a
number of obstetric and neonatal complications in the setting of PAS, including
birth asphyxia, preeclampsia, chorioamnionitis, cardiac anomalies, polycythemia,
and systemic infection.7,12- 15 Others
have failed to find a significant difference in the frequency of perinatal
complications between infants with PAS and controls.16 Hematologic
disorders, including factor V Leiden mutation and hyperhomocysteinemia, may
also play a role in the pathogenesis of PAS.14,17- 19
Previous studies of PAS are subject to a number of important limitations.
Most describe only a small number of children7,12,16,20- 22 or
lack an adequate comparison group.8,14,23 We
found previously that preeclampsia and intrauterine growth restriction are
independent risk factors for PAS.15 However,
our earlier study included only children with PAS who developed long-term
motor impairment, and we did not confirm the diagnosis of PAS with review
of brain imaging. Therefore, we set out to determine pregnancy complications
associated with radiologically confirmed PAS in a defined population of infants.
This case-control study was nested within the cohort of all 199 176
infants born from January 1, 1997, to December 31, 2002, in the Northern California
Kaiser Permanente Medical Care Program (KPMCP). The KPMCP is a large managed
care organization that provides care for more than 30% of the population in
northern California. The members of KPMCP are demographically similar to the
California population, except that the very poor and very wealthy are underrepresented.24 All study procedures were approved by the institutional
review boards at KPMCP and at the University of California, San Francisco,
which waived requirement for informed consent.
We searched electronically all head magnetic resonance imaging (MRI)
and computed tomographic (CT) reports generated from January 1, 1997, through
July 1, 2003 (all study infants were all at least 6 months of age when records
were searched), and retrieved neuroimaging reports containing any of the following
text strings: stroke, infarct, thrombo, ischemi, middle cerebral artery, MCA, posterior cerebral artery, PCA, anterior cerebral artery, ACA, vascular insult, vascularinjury, vascularevent, porencephal, or hydranencephal.
We also retrieved all head MRI and CT reports generated for newborns admitted
to the neonatal intensive care unit with seizures and for children who were
given an inpatient or outpatient physician diagnosis of stroke, developmental
delay, hemiparesis, or cerebral palsy before November 2003.
Two study investigators (K.H.B., J.L.) independently reviewed the head
MRI and CT reports of 1970 patients to identify potential cases of PAS. They
excluded neuroimaging studies that revealed (1) no mention of possible stroke;
(2) an acute infarction occurring after 28 days of life; (3) isolated hemorrhagic
lesions; (4) venous infarction; or (5) infarctions limited to the arterial
watershed zones.1 A study neuroradiologist
(C.L.) then reviewed 141 head MRI or CT scans to confirm the presence of an
arterial-distribution ischemic infarction. When the presence of an arterial
infarction was not obvious, the imaging studies were reviewed by a second
neuroradiologist (A.J.B.), and a consensus about the final neuroimaging diagnosis
was reached through discussion. One child whose MRI scan could not be located
was included because his MRI report described an unambiguous large posterior
cerebral artery infarction.
Two investigators (K.H.B., J.L.) reviewed the medical records of children
with radiologically confirmed arterial infarction to determine whether the
stroke was perinatal. We defined a perinatal event
as one that occurred in utero or up to 28 days after birth.25 If
the infant presented within the first 28 days after delivery, the stroke was
considered to be acute. If an infant had been considered
neurologically normal before 1 month of age but was later diagnosed with an
old arterial infarction, the stroke was considered to be presumed perinatal.3
Three infants with PAS caused by a known neonatal event were excluded
from the risk factor analysis but included in the prevalence calculation.
One infant had an iatrogenic PAS after surgical evacuation of an intracranial
hematoma, and 2 had acute arterial strokes in the setting of severe meningitis
after 2 weeks of age.
We randomly selected 3 controls per case from the study population.
Control infants were frequency matched to the infants with PAS on birth year,
facility of birth, and gestational age stratum (<32 weeks, 32-35 weeks,
and ≥36 weeks of gestation).
Two study investigators blinded to case status reviewed prenatal (J.L.),
obstetric (J.L.), and neonatal (L.H.H.) medical records by using a standardized
protocol. An infant born at or later than 41 weeks of gestation was considered
postdates. Maternal body mass index was calculated from prepregnancy height
and weight measurements, if available, or from the first prenatal visit. To
evaluate the effect of ethnicity on risk of PAS, we abstracted maternal ethnicity
according to self-report as noted in the medical records. Intrauterine growth
restriction was defined as birth weight less than the 10th percentile for
gestational age according to race- and sex-specific normative data compiled
from California births.26 The mother was considered
to have a history of infertility if this was documented in a prenatal, obstetric,
or neonatal record. Information about the use of infertility drugs for the
index pregnancy was also abstracted from these records.
Preeclampsia was defined as a physician diagnosis of either preeclampsia
or pregnancy-induced hypertension. We used the term chorioamnionitis to indicate a maternal temperature of at least 37.8°C or a physician
diagnosis of chorioamnionitis according to clinical symptoms alone. The term cord abnormalities included tight nuchal cord, umbilical
cord knot, and body cord. We used the term birth asphyxia to indicate a diagnosis made by a treating physician of either birth
asphyxia or hypoxic-ischemic encephalopathy. The second stage of labor was
coded as prolonged if it lasted more than 2 hours. Fetal heart rate abnormalities
were considered present if a treating physician noted repetitive or prolonged
late decelerations, fetal bradycardia, nonreassuring fetal heart tracing,
or fetal distress according to electronic fetal heart rate monitoring. Decreased
fetal movement referred to a maternal report of decreased fetal movement before
labor or decreased fetal movement noted during a nonstress test.
We compared dichotomous variables by using χ2 or Fisher
exact test and continuous variables by using the t test.
We calculated univariate odds ratios (ORs) and 95% confidence intervals (CIs)
with the Cornfield or exact method, as appropriate, and multivariate ORs with
backward stepwise logistic regression, with P<.10
used as the cutoff for retention in the model.27 To
determine whether risk factors differed for the acute-presentation group compared
with the delayed-presentation group, we performed a polytomous logistic regression.27 Odds ratios closely approximate the relative risk
because the outcome of PAS is rare.
The multivariate model included maternal (primiparity and infertility),
prepartum (preeclampsia, oligohydramnios), and intrapartum (chorioamnionitis,
prolonged rupture of membranes, cord abnormality, and use of oxytocin) characteristics
associated with PAS at a level of P<.15 in univariate
analyses that were considered unlikely to be a result of the stroke. Infant
characteristics such as neonatal seizures and low Apgar scores, as well as
decreased fetal movement, were not included in the model because they most
likely result from perinatal brain injury and do not play a causal role.
It is impossible to know where some variables lie on the causal pathway.
For instance, fetal heart rate abnormalities may (1) play a primary causal
role in PAS; (2) be on the causal pathway between a preceding risk factor
and PAS; (3) be merely an adverse effect of an underlying causal factor without
direct impact on PAS; (4) be a direct consequence of the stroke event; or
(5) be a combination of the above. For this reason, intrapartum complications
associated with PAS, including fetal heart rate abnormalities, prolonged second
stage of labor, vacuum assistance, and emergency cesarean delivery, were not
included in the main multivariate analysis but instead were added to the model
separately to determine whether they contribute additional risk to PAS beyond
that accounted for by the variables in the main model.
Among 1970 children who had either a head-imaging report or a clinical
diagnosis suggesting possible PAS, 141 patients had a head MRI or CT scan
that was reviewed by a neuroradiologist (Figure). A total of 40 cases of PAS were confirmed, providing a population
prevalence of 20 per 100 000 live births. Infants with PAS were all singleton
gestation, and the majority (85%) were delivered at term (mean [SD], 39.9 [1.0]
weeks). The 6 preterm infants with PAS were born between 30 and 35 weeks of
gestation, with the exception of 1 infant who was delivered at 24 weeks.
Most infants with PAS (58%) presented during the acute neonatal period.
Term infants presented frequently with neonatal seizures (70%), whereas 3
of 4 preterm infants with acute PAS were diagnosed incidentally when a routine
head ultrasound showing intraventricular hemorrhage or white-matter abnormalities
led to a head CT that diagnosed an arterial stroke. All children with presumed
perinatal stroke presented after 2 months of age, with pathologic handedness
(hand preference earlier than 1 year of age) as the most common presenting
symptom, which was consistent with previous reports.3
Previously described causes of PAS were uncommon in our cohort. Although
cardiac echocardiography was not performed routinely, none of the children
with PAS were diagnosed with a major congenital heart abnormality. Minor cardiac
findings in 4 patients with a heart murmur included a patent ductus arteriosus,
a very mild muscular hypertrophy, and a small ventral septal defect that was
thought to be clinically insignificant. One infant had a mild polycythemia
(hematocrit level dropped from 70.5% to 65% during 20 hours) that did not
require exchange transfusion. Few patients received a coagulation evaluation,
and therefore we could not assess the role of prothrombotic disorders in PAS.
The diagnosis of PAS was made on either head MRI (70%) or head CT (30%).
Unilateral infarctions were more common on the left (53%) than on the right
(35%), whereas 13% demonstrated bilateral arterial distribution infarcts.
The majority of strokes involved only the middle cerebral artery distribution
(74%), with an additional 4 strokes that involved the middle cerebral artery
plus other arteries.
Although maternal age, race, body mass index, and number of previous
miscarriages did not differ between the case and control groups, case mothers
were more likely to be primiparous (Table 1).
Case mothers were also more likely to have a history of infertility, although
the difference was not significant (11% vs 4%, P = .11).
Other prepartum characteristics more commonly observed in case mothers included
preeclampsia, oligohydramnios, and decreased fetal movement.
Intrapartum complications associated with PAS included chorioamnionitis,
prolonged rupture of membranes, prolonged second stage of labor, fetal heart
rate abnormality, cord abnormality, vacuum assistance, and emergency cesarean
delivery. A clinical diagnosis of birth asphyxia or hypoxic ischemic encephalopathy
was given to 6 infants with PAS (16%), whereas none of the control infants
received this diagnosis (P<.001). After delivery,
infants with PAS were significantly more likely to be given an Apgar score
less than 7 at 5 minutes and to require resuscitation (Table 1). The umbilical artery pH was less than 7.0 in 3 of 11 infants
with PAS for whom a cord gas result was available.
Chorioamnionitis was significantly associated with PAS only in the absence
of fetal distress (OR, 5.4; 95% CI, 1.3-21.6). No other interactions were
found in stratified analyses. Furthermore, chorioamnionitis was associated
with acute stroke (OR, 6.1; 95% CI, 1.9-19.2) but not with presumed perinatal
stroke (OR, 0.7; 95% CI, 0.1-6.0). The risk factors for acute and presumed
perinatal stroke were otherwise similar.
The following variables were entered into the main logistic regression
model: history of infertility, oligohydramnios, preeclampsia, prolonged rupture
of membranes, cord abnormality, chorioamnionitis, primiparity, and use of
oxytocin (Table 2). The risk factors
that remained independently associated with PAS were infertility (OR, 7.5;
95% CI, 1.3-45), preeclampsia (OR, 5.3; 95% CI, 1.3-22.0), chorioamnionitis
(OR, 3.4; 95% CI, 1.1-10.5), and prolonged rupture of membranes (OR, 3.8;
95% CI, 1.1-12.8). Seven of the 8 infertile women received ovarian-stimulation
medications before the conception of the index child. When ovarian stimulation
was entered into the model in the place of infertility, the adjusted OR was
13.2 (95% CI, 1.8-98.3).
Although fetal heart rate abnormalities, vacuum extraction, and emergency
cesarean delivery were all significantly associated with increased risk of
PAS in univariate analyses, none of these factors remained independently associated
with PAS after adjustment for the other variables in the multivariate model.
Prolonged second stage of labor was a significant univariate predictor of
PAS but was also strongly correlated with primiparity. As expected, when primiparity
was removed from the model to avoid collinearity, prolonged second stage of
labor was an independent predictor of PAS (OR, 5.0; 95% CI, 1.2-21.1; P = .03).
The following factors were significantly associated with PAS on either
univariate or multivariate analysis and could be identified before delivery:
primiparity, infertility, oligohydramnios, preeclampsia, chorioamnionitis,
prolonged rupture of membranes, decreased fetal movement, prolonged second
stage of labor, and fetal heart rate abnormalities. As expected, the risk
of PAS increased with the number of these risk factors present (Table 3). Only 6% of controls had 3 or more risk factors present
compared with 60% of case children (OR, 25.3; 95% CI, 7.9-87.1). When 3 or
more risk factors are present, the probability of delivering a child with
PAS is as high as 1 in 200.
Only 3 placental pathologic examinations were performed in the 40 infants
with PAS. Findings included a positive staphylococcus culture on the fetal
side (1 term infant), funisitis (1 preterm infant), and acute chorioamnionitis,
together with a placental infarction (1 preterm infant). Eleven control placentas
were submitted for pathologic examination. Findings among the 3 control infants
born at term included a chorangioma (1), placental abruption (1), and a normal
placenta (1), whereas for the 8 preterm control infants, findings included
acute chorioamnionitis (3) and a normal placenta (5).
Perinatal arterial stroke is the most common cause of hemiplegic cerebral
palsy, yet the etiology is poorly understood. To our knowledge, this is the
first controlled study of risk factors for PAS within a population that includes
all cases diagnosed by neuroimaging, and we found several significant risk
factors identified during pregnancy. Intrapartum complications were also more
common in infants with PAS, and the risk of PAS was dramatically higher in
the presence of multiple risk factors.
The prevalence of PAS has not been clearly determined. Previous population-15,28,29 and hospital-based14,16,30 estimates range from
17 to 93 per 100 000 live births, depending on the study design and case
definition. We found that PAS was diagnosed in 20 per 100 000 live births.
The exclusion of periventricular and watershed distribution infarctions, as
well as infarctions that appear to be venous in origin, may explain why our
prevalence estimate is within the lower range of previous estimates. Our findings
confirm, however, that the rate of PAS is 17 times higher than the incidence
of childhood ischemic stroke31 and as high
as the annual incidence of large-vessel ischemic stroke in adults older than
18 years (17-23 per 100 000).32
Although infertility did not represent a significant risk factor in
univariate analysis, it was an independent predictor of PAS after adjustment
for confounders. This discrepancy is explained by the fact that infertile
women were less likely to have other significant risk factors such as preeclampsia
and oligohydramnios, and this negative confounding was largely removed by
multivariate analysis. All 4 mothers with infertility who had a child with
PAS in our study were treated with ovarian-stimulation drugs. Ovarian hyperstimulation
syndrome is known to cause a hypercoagulable state that may lead to arterial
and venous thrombosis in the mother.33,34 To
our knowledge, this is the first report of infertility treatment with ovarian
stimulation associated with thrombotic complications in the newborn; whether
placental thrombosis underlies this relationship requires further study. Others
have reported an increased risk of cerebral palsy after infertility treatment.35,36 Larger studies are needed to confirm
the association between infertility and PAS and to determine whether the underlying
infertility or its treatment might be responsible.
Preeclampsia, a strong risk factor for PAS in our population, is thought
to result from a vascular defect in the placenta bed, leading to reduced uteroplacental
blood flow.37,38 Preeclampsia
has been associated with maternal prothrombotic disorders, thrombotic lesions
in the placenta, and a maternal history of thromboembolism.39- 41 Furthermore,
preeclampsia increases the risk of neonatal sinovenous thrombosis, neonatal
encephalopathy, and motor impairment related to PAS.15,42- 44 The
relationship between prothrombotic disorders, preeclampsia, placental dysfunction,
and PAS deserves further study.
Chorioamnionitis, a diagnosis that is usually based on the presence
of maternal intrapartum fever, has received much attention as a predictor
for cerebral palsy.45 However, to our knowledge,
this is the first report of clinical chorioamnionitis acting as an independent
risk factor for PAS. Interestingly, chorioamnionitis was associated only with
PAS that was diagnosed in the neonatal period and not with PAS diagnosed later
in infancy, which suggests that the pathogenesis of PAS may differ for infants
with earlier presentation compared with those with a later presentation. Alternatively,
the presence of chorioamnionitis may increase the likelihood of neonatal symptoms,
thus prompting earlier recognition of PAS.
We found that intrapartum complications were more common in infants
with PAS than in control infants, which was consistent with previous reports.20,21,46 However, complications
such as fetal heart rate abnormalities were no longer independently associated
with PAS after adjustment for maternal variables. Furthermore, although fetal
distress and low Apgar scores often lead to a clinical diagnosis of birth
asphyxia, these complications do not always reflect a global hypoxic-ischemic
event, as implied by the term birth asphyxia. In
our study, all 6 infants with PAS who were diagnosed with birth asphyxia had
a focal arterial infarction as opposed to the more typical neuroimaging findings
of hypoxic-ischemic brain injury, such as deep gray-matter or arterial-watershed
injury, reminding us that the clinical diagnosis of birth asphyxia is not
specific for any single pathogenetic mechanism of brain injury.
Our study was subject to a number of limitations. Despite the large
source population, the number of final cases was relatively small, resulting
in wide CIs for several effect estimates. There may be underascertainment
of PAS because 11% of our cohort was lost to follow-up by 1 year of age, and
children with subtle neurologic deficits may not have received a neuroimaging
study to identify a PAS. Insufficient information was available on possible
risk factors such as smoking, drug use, and socioeconomic indicators. Given
the retrospective nature of the study, we lacked a uniform neuroimaging protocol
and relied on medical record documentation for determining the presence of
risk factors. However, physicians typically document maternal complications
in the medical records before knowing the outcome of the child, thus minimizing
potential reporting bias. Strengths of our study include the population-based
setting; blinded evaluation of maternal, prepartum, and intrapartum factors;
and the selection of an appropriate control group.
Perinatal arterial stroke has been attributed to etiologies such as
congenital heart disease,47 polycythemia,13 in utero cocaine exposure,48 and
neonatal meningitis.49 Among our cases, only
2 infants had neonatal meningitis, 1 had an iatrogenic stroke from brain surgery,
no infants had major congenital heart disease, and 1 infant had a mild polycythemia
that resolved without intervention. If one makes an assumption that the 4
independent risk factors found in this study do in fact play a causal role,
the population-attributable fractions (chorioamnionitis, 21%; infertility,
19%; preeclampsia, 16%; and prolonged rupture of membranes, 16%) suggest that
these factors may account for a significant proportion of PAS cases.
A thromboembolic event originating from an intracranial vessel, extracranial
vessel, the heart, or the placenta may lead to an arterial stroke that occurs
in the perinatal period.1 Although placental
findings have rarely been described in PAS, pathologic studies have suggested
that placental infarction and inflammation may be related to cerebral palsy,
a long-term outcome of PAS.50- 53 We
found similar placental abnormalities in our infants with PAS, although only
3 placentas were examined. Maternal conditions that affect placental function,
such as preeclampsia and chorioamnionitis, were found to be independent risk
factors for PAS. Together, these data suggest that the placenta may play an
important role in the pathogenesis of PAS, although no conclusions can be
drawn from this study, given the small number of placentas examined.
Future studies that provide additional data on placental pathology,
coagulation disorders in the mother and fetus, and comprehensive neuroimaging
findings may further elucidate the underlying pathogenetic mechanisms of PAS.
Only with an improved understanding of what causes PAS will we determine how
to prevent this important cause of neurologic disability in children.
Corresponding Author: Yvonne W. Wu, MD,
MPH, UCSF Department of Child Neurology, Box 0137, 350 Parnassus Ave, Suite
609, San Francisco, CA 94117 (email@example.com).
Author Contributions: Dr Wu had full access
to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Croen, Backstrand,
Ferriero, Fullerton, Wu.
Acquisition of data: Lee, Croen, Backstrand,
Yoshida, Henning, Lindan, Barkovich, Wu.
Analysis and interpretation of data: Lee, Croen,
Backstrand, Ferriero, Wu.
Drafting of the manuscript: Lee, Backstrand,
Critical revision of the manuscript for important
intellectual content: Croen, Backstrand, Yoshida, Lindan, Ferriero,
Fullerton, Barkovich, Wu.
Statistical analysis: Lee, Wu.
Obtained funding: Ferriero, Wu.
Administrative, technical, or material support:
Lee, Croen, Yoshida, Henning, Wu.
Study supervision: Croen, Barkovich, Wu.
Financial Disclosures: None reported.
Funding/Support: This study was funded by the
United Cerebral Palsy Foundation. Dr Wu was supported in part by NS35902 from
the National Institutes of Health and the National Institute of Neurological
Disorders and Stroke, and Ms Lee was supported by the Doris Duke Clinical
Research Training Program.
Role of the Sponsor: The United Cerebral Palsy
Foundation had no role in the design and conduct of the study; collection,
management, analysis, and interpretation of the data; or in the preparation,
review, or approval of the manuscript.
Acknowledgment: We thank Carla Stelling, BS,
Roxana Odouli, MSPH, Rick Riordan, BS, and Kim Detels for their research assistance,
as well as Karin Nelson, MD, S. Claiborne Johnston, MD, PhD, and Thomas B.
Newman, MD, MPH, for their mentorship and careful reviews of the manuscript.