OBJECTIVE. Few data exist regarding rates and predictors of recurrence after childhood arterial ischemic stroke. We sought to establish such rates within a large, multiethnic population and determine whether clinical vascular imaging predicts recurrence.
PATIENTS AND METHODS. In a population-based cohort study, we collected data on all documented cases of arterial ischemic stroke among 2.3 million children (<20 years old) enrolled in a northern Californian managed care plan from January 1993 to December 2004. Perinatal strokes were those that occurred by 28 days of life. Data on cerebrovascular imaging (conventional or magnetic resonance angiography), including presence of vascular abnormalities, were abstracted from official radiology reports. We used Kaplan-Meier survival-analysis techniques to determine rates and predictors of recurrent stroke.
RESULTS. Among 181 incident childhood stroke cases (84 perinatal; 97 later childhood), there were 16 recurrent strokes (1 after a perinatal stroke) at a median of 2.7 months. The 5-year cumulative recurrence rates were 1.2% after perinatal stroke and 19% after later childhood stroke. Of the 97 children with later childhood strokes, 52 received cerebrovascular imaging, predominantly magnetic resonance angiography (n = 36) and conventional angiography (n = 26). Although there were no recurrences among children with normal vascular imaging, children with a vascular abnormality had a 5-year cumulative recurrence rate of 66%.
CONCLUSIONS. Strokes recur in one fifth of cases of later childhood arterial ischemic stroke but are rare after perinatal stroke. Among the later childhood cases, cerebrovascular imaging identifies those at highest risk for recurrence.
Despite increasing recognition of arterial ischemic stroke (AIS) as an important cause of childhood disability, few data exist regarding rates and predictors of recurrent childhood stroke.1,2 Although children with perinatal strokes are thought to be at lower risk than older children, a direct comparison of perinatal versus later childhood strokes within the same population has never been made.1,3,4 Stroke etiology also likely impacts recurrence risk. Conditions associated with first childhood stroke include congenital heart disease, sickle cell disease, meningitis, a variety of hereditary and acquired prothrombotic states, and a number of vasculopathies.5–8 In a study of children identified in several German hospitals, children whose strokes were classified as “vascular” in etiology had fourfold odds of suffering a recurrence compared with children with idiopathic strokes,2 but this finding has not been validated in a distinct clinical setting or a more diverse ethnic population.
The dearth of information on recurrence after childhood stroke likely contributes to practice variability in the evaluation and treatment of childhood stroke. Several paragraphs embedded within the American College of Chest Physicians' guidelines on antithrombotic therapy in children comprise the sole in-print guidelines addressing childhood AIS.9 There are no established standards of care for the diagnostic evaluation of children with stroke, and head and vascular imaging are used variably.10 Furthermore, secondary prevention practices vary widely, and many children receive no specific intervention to reduce risk of recurrence.10–12
We sought to study the risk of stroke recurrence among unselected children and neonates in a defined multiethnic population. The objectives of this study were to determine (1) recurrence rates after childhood AIS during long-term follow-up, (2) the relative risk of recurrence in children with perinatal versus later childhood AIS, and (3) the utility of clinical cerebrovascular imaging in predicting recurrence risk.
After obtaining the approval of the Kaiser Permanente Medical Care Program (KPMCP) and University of California, San Francisco, institutional review boards, we performed a retrospective cohort study in which we sought to identify all strokes occurring within the population of children (<20 years of age) enrolled in KPMCP during an 11-year study period (January 1993 through December 2003). This report describes only the children with AIS.
KPMCP is the largest nonprofit managed care organization in the country, with 16 hospitals and 36 outpatient facilities in northern California. It provides typically long-term care for ∼3.1 million people, or 30% of the regional population. The KPMCP population shares the general sociodemographic distribution of California except for an underrepresentation of socioeconomic extremes.13
Cases were ascertained through a multitiered process. First, we electronically searched the KPMCP patient databases for hospital discharge diagnoses (coded by a medical charts abstractor) and outpatient diagnoses (coded by the treating physician) suggestive of a stroke. This included diagnoses of ischemic stroke (International Classification of Diseases, Ninth Revision [ICD-9] codes 433–436, 437.6, and 325), subarachnoid hemorrhage (ICD-9 codes 430 and 772.2), and intracerebral hemorrhage (ICH; ICD-9 code 431). The search included all out-of-plan hospitalizations. Second, we performed keyword searches of all electronic head imaging reports (MRI and computed tomography scan) with the following text strings: stroke, infarct, infarction, infarcted, thrombus, thromboembolic, thromboembolism, thrombotic, thrombosis, ischemia, ischemic, lacune, lacunar, vascular event, porencephaly, and porencephalic. A single pediatric stroke neurologist (Dr Fullerton) reviewed all identified reports. Third, we cross-referenced with previous studies of cerebral palsy and perinatal arterial ischemic stroke that used a KPMCP birth cohort (January 1991 through December 2002).4,14
Potential cases were subjected to chart review. Two child neurologists (Drs Fullerton and Wu) independently confirmed cases of childhood stroke; a third stroke neurologist (Dr Johnston) arbitrated disputes. The criteria for AIS were: (1) documented clinical presentation consistent with stroke, such as a sudden onset focal neurologic deficit; and (2) computed tomography scan or MRI showing a focal ischemic infarct in a location and of a maturity consistent with the neurologic signs and symptoms. Cases were excluded if the stroke occurred before the child's enrollment in Kaiser or outside of the study period.
Perinatal strokes were those occurring between 28 weeks' gestation and 28 days of life.3,4 They included presumed perinatal strokes, those that presumably occur perinatally but are not diagnosed until the children become symptomatic later in life.15 Strokes occurring after 28 days were called later childhood strokes.
A single pediatric nurse medical charts analyst reviewed records of all confirmed cases by using a standardized protocol. Dr Fullerton rereviewed all records and used all available data to categorize the stroke etiology: cardiac (including congenital and valvular heart disease), infection (including meningitis, encephalitis, and sepsis), hypercoaguable state (acquired or hereditary), primary vasculopathy (arterial dissection, moyamoya [a progressive occlusive vasculopathy of the intracranial blood vessels], vasculitis, and arterial stenosis in the absence of another etiology, such as meningitis), other, or idiopathic.2,8
We abstracted data regarding any cerebral vascular imaging (magnetic resonance angiography, computed tomography angiography, or catheter angiography) by using the official interpretation by the clinical radiologist. We defined “abnormal vascular imaging” as any “stenosing” vascular abnormality: arterial stenosis, moyamoya, arterial dissection, fibromuscular dysplasia, or vasculitis.2,8 Our definition excluded isolated arterial occlusion, which is more likely to represent thrombus than a vasculopathy.2
Recurrent Cerebrovascular Events
KPMCP members receive all routine care within the system, facilitating the acquisition of follow-up data. Out of plan services, including emergent admissions to outside hospitals, are recorded in an electronic database, and copies of medical charts from outside admissions are typically available in outpatient charts. We abstracted data regarding any additional strokes or transient ischemic attacks (TIAs) that occurred after the index AIS, applying the same adjudication procedures but blinding the adjudicators to the vascular imaging reports. Because children with certain ischemic stroke etiologies (eg, moyamoya) are also at risk for hemorrhagic stroke, hemorrhagic strokes after the index AIS were considered recurrences. Similar clinical and radiologic criteria were used for this stroke subtype. Hemorrhagic transformation of the initial infarct was not considered a recurrence. Criteria for the diagnosis of TIA included (1) a focal neurologic deficit of acute onset lasting <24 hours, (2) no radiographic evidence of an infarct, and (3) clinical suspicion of a TIA by a physician.
Incidence rates were calculated as the number of strokes divided by the number of person-years at risk. We used survival analysis to determine recurrence rates. Our primary outcome variable was time to recurrent stroke; for this analysis, the period at risk began on the date of the index AIS and ended on the date of the first recurrent stroke (the “failure” event) or censoring. Cases were censored (ie, withdrawn from the survival analysis) at either death or loss to follow-up by using the date of the last recorded visit to a KPMCP facility. Our primary outcome variable was time to recurrent stroke defined as date of the index AIS to date of the first recurrent stroke. Cumulative recurrence rates were derived from hazard functions. A secondary analysis used the combined outcome of stroke or TIA.
We constructed Kaplan-Meier survival curves to compare recurrence-free survival rates between different subgroups and used log-rank tests to determine the significance of univariate associations (α set at .05). Univariate Cox proportional hazards regression techniques were used to assess relative risk in terms of hazard ratios (HRs) with 95% confidence intervals (CIs). We used log-minus-log survival plots to demonstrate proportionality for binary predictors.16 However, we primarily used techniques that are not dependent on the proportional hazards assumption: comparisons of cumulative recurrence rates and log-rank tests.
χ2 tests were used to compare simple proportions. We used the nonparametric test for trends across ordered groups to determine whether the proportion of children receiving vascular imaging changed over time.17 We used Stata 9.0 (Stata Corp, College Station, TX) to perform all statistical calculations.
Our study of 8963308 person-years included 2347982 individual children enrolled in KPMCP for a mean of 3.8 years during the 11-year study period. A total of 181 cases of childhood AIS were confirmed, yielding an overall annual incidence of 2.0 per 100000 person-years. Of these, 84 (46%) were perinatal; 97 (54%) occurred in later childhood. The vast majority (98%) were first-time strokes; only 3 children had previous strokes (before the study period or before enrollment in KPMCP). The stroke cohort was ethnically diverse (16% black, 46% non-Hispanic white, 19% Hispanic, 12% Asian, 1% Native American, and 6% other or unknown), with a preponderance of boys (55%). Etiologies of the index strokes are shown in Table 1. Antithrombotic agents were used after 2% of perinatal strokes and after 51% of later childhood strokes (Table 1).
A total of 64 children had cerebrovascular imaging; such imaging was uncommonly performed after perinatal stroke (Table 2). Compared with the 52 older children who received vascular imaging, the 46 who did not receive vascular imaging were more likely to have either a cardiac etiology (22% vs 4%; P = .006) or infectious etiology (36% vs 12%; P = .005). There was no change in rates of vascular imaging over the 11 years of study (P = .91).
Follow-up data were available for 84 (100%) children with perinatal strokes and 92 (95%) children with later childhood strokes. Only 6 children died during the acute period (2 neonates). The perinatal stroke group was followed for a median of 5.8 years (mean: 6.0 years [range: 3 days to 12.4 years]), compared with a median of 5.1 years (mean: 5.0 years [range: 5 days to 11.8 years]) for the later childhood stroke group.
We identified a total of 16 subjects with first recurrent strokes occurring at a median of 2.2 months after the index stroke (range: 1 day to 4 years). Fourteen recurrences were AIS and 2 were ICH; only 1 (an ICH) occurred after a perinatal stroke. Etiologies of the first recurrent events are shown in Table 1. Six recurrences (all AIS) were in children treated with antithrombotic agents (3 received aspirin, 2 low molecular weight heparin, and 1 warfarin), whereas 10 (including the 2 ICH) were in children who received no antithrombotic therapy (Table 1). Although hemorrhagic transformation of the initial AIS was not considered a recurrent event, there was no ICH among the children treated with antithrombotic agents.
Recurrence After Later Childhood Stroke
The cumulative stroke recurrence rate was 15% (95% CI: 12%–30%) at 1 year, and 19% (95% CI: 12%–30%) at 5 years (Fig 1). Eleven of the 15 recurrences occurred within the first 6 months, and only 2 occurred beyond 1 year. The single ICH after a later childhood stroke was in a child with systemic lupus erythematosis who subsequently had 2 recurrent AISs 4 months after the ICH. There were 4 additional children who had recurrent TIAs without recurrent strokes. The 5-year cumulative recurrence rate for any cerebrovascular event (stroke or TIA) was 24% (95% CI: 16%–36%). Of the 19 subjects with recurrent cerebrovascular events, 11 had multiple recurrences, with 6 having >3 recurrences.
Recurrence After Perinatal Stroke
The single recurrence after perinatal stroke was in a neonate with meningoencephalitis who suffered an ICH 7 days after a small internal capsule AIS. Overall, children with perinatal strokes were 94% less likely to suffer a recurrent stroke than those with later childhood strokes (HR: 0.06; 95% CI: 0.008–0.48; P = .008). Conversely, children with later childhood strokes had a 16-fold increased hazard of recurrence compared with the neonates (HR: 16; 95% CI: 2.1–120; P = .008).
Abnormal Vascular Imaging Predicts Recurrence
We limited this analysis to the higher-risk later childhood subgroup. Of the 15 recurrent strokes, 13 occurred among children who had received vascular imaging. Abnormal vascular imaging predicted recurrence (Fig 2). None of the children with normal vascular imaging had a recurrent stroke, whereas two thirds of those with vascular abnormalities suffered a recurrence within 5 years (Table 3).
Receiving vascular imaging was itself associated with an increased risk of recurrence (HR: 5.9; 95% CI: 1.3–26; P = .02), although this association was not significant after excluding 5 children whose initial vascular imaging was performed after their first recurrent event (HR: 3.9; 95% CI: 0.82–18; P = .09). To account for possible selection bias affecting the association between abnormal vascular imaging and recurrence, we excluded these 5 children (3 of whom had additional recurrences after the vascular imaging) from the primary analysis. The remaining children with abnormal vascular imaging had a 5-year cumulative recurrence risk of 57% (95%: CI 33%–84%), again significantly greater than children with normal vascular imaging (P < .0001 by log rank). To assess the possible impact of differential treatment with antithrombotic agents, we compared treatment rates and found them to be similar among children with abnormal (71%) versus normal (77%) vascular imaging (P = .63 by χ2 test). Furthermore, treatment with an antithrombotic agent did not predict recurrence (P = .42 by log rank).
In an unselected sample from a defined, multiethnic population followed for several years, we found that (1) recurrence rates after childhood AIS are high in an easily identified subgroup, those older than 28 days, (2) the first 6 months are the highest risk period for recurrence, and (3) clinical cerebrovascular imaging is useful for prognostication. Because most children (>28 days old) in our cohort were treated with antithrombotic agents, the true natural history recurrence rates are likely even higher than those that we observed.
Cerebrovascular imaging, although commonly performed in the evaluation of an adult stroke patient, often is not performed in children with AIS. Slightly under half of the cases in our study had such imaging, and there was no change in rates of vascular imaging over the 11 years of study. The only in-print guidelines for the management of later childhood AIS do not explicitly recommend cerebrovascular imaging.9 However, our data demonstrate that these imaging studies can identify those children at highest risk for recurrence: by 5 years, recurrent strokes occurred in none of the children with normal vascular imaging, compared with almost 70% of children with vascular abnormalities. Furthermore, because we based our definition of abnormal vascular imaging on the interpretations of clinical radiologists (many of whom have no formal training in neuroradiology), this finding is generalizable to the typical hospital setting.
Arterial stenosis was the most commonly identified vascular abnormality in our study. It was an isolated finding in 64% of children with stenosis and occurred in the setting of a more diffuse vasculopathy (moyamoya or vasculitis) in the remainder. Although the radiographic appearance of a narrowed vessel lumen can result from a vasculopathy or a recanalizing embolus, the high recurrence risk that we observed with this lesion, similar to primary vasculopathies such as moyamoya, suggests that arterial stenosis behaves more like a true vasculopathy in terms of recurrence. The etiology of isolated large vessel stenosis in children remains unknown and may be heterogeneous. These lesions could be congenital, such as the hypoplastic arteries seen in PHACE (posterior fossa malformations, hemangiomas, arterial anomalies, coarctation of the aorta, cardiac defects, and eye abnormalities) syndrome,18,19 or acquired. Many of these cases may represent an entity termed “transient cerebral arteriopathy,” a monophasic focal arteriopathy of childhood involving the large vessels, most often the distal internal carotid artery. However, this diagnosis requires follow-up vascular imaging demonstrating nonprogression, and such imaging was uncommon in our study.20,21 Regardless, transient cerebral arteriopathy is a provisional diagnosis that does not specify a pathophysiology, although an infectious etiology (specifically varicella zoster virus) is suspected.21,22
Our definition of abnormal vascular imaging excluded isolated arterial occlusion because such a finding could be present in the absence of a true vasculopathy, such as after a cardioembolic event.2 It is possible that some children with isolated occlusion had an underlying vasculopathy and were, therefore, misclassified into the “normal” vascular imaging group. Misclassification of a high-risk lesion into the “normal” group could have biased our results toward the null; however, there were no recurrences among the 6 children with isolated arterial occlusion, indicating that no such bias occurred.
Identification of a child at high risk for recurrence on the basis of vascular imaging not only provides prognosis, but also an opportunity for risk reduction. Although there have been no randomized, controlled trials of secondary stroke prevention in children, a variety of strategies are used to reduce the likelihood of a recurrent stroke in the setting of specific high-risk vascular abnormalities. Examples are anticoagulation for arterial dissections,9 vascular bypass procedures for moyamoya,23 and immunosuppressive agents for autoimmune vasculitides.24,25
Only 2 previously published studies performed hazard analyses of recurrence after childhood stroke, and both used selected populations in Germany. In a study of perinatal stroke, using a convenience sample and a broader definition of recurrence (including venous sinus thrombosis and deep venous thrombosis), the 5-year cumulative recurrence rate was 3.5%, within the 95% CIs of that found in our study.1 A study of later childhood stroke, which used a sample derived from hospital populations in various geographic regions of Germany, reported a recurrent stroke rate of 5% within 5 years, which was below the lower limit of our 95% CIs, although their CIs (not reported) may overlap ours.2 A true difference, however, could be attributable to their lower reported prevalence of vasculopathies (18%).
Our study's greatest limitation was that vascular imaging was performed only on the subgroup of children whose treating physicians deemed it necessary. Children who did not receive vascular imaging were more likely to have a cardiac or infectious stroke etiology, suggesting that treating physicians may not have pursued vascular imaging once an explanation for the stroke was obtained. Among those who received vascular imaging, the results clearly allowed stratification into high- and low-risk groups. However, repeat vascular imaging was seldom performed, and recent data suggest that progression of cerebrovascular abnormalities may further stratify recurrence risk in children.26 A additional limitation is that we cannot comment on the preferred type of vascular imaging in children. However, given the high yield of magnetic resonance angiography in children with AIS, as well as the lower risks associated with this noninvasive study, magnetic resonance angiography is probably a reasonable first step.27,28
The strengths of our study include the multiethnic, population-based setting enhancing generalizability, the relatively large sample size given the rarity of this disease, and the extent and quality of follow-up data. In addition, by basing our definition of vascular imaging abnormalities on formal radiology reports, rather than reinterpretation by a study investigator, we were able to demonstrate the prognostic value of such imaging in a typical hospital setting.
We conclude that children with later childhood AIS are at high risk for recurrent stroke, particularly within the first 6 months, and cerebrovascular imaging identifies those children at highest risk. Cerebrovascular imaging should be included in the diagnostic evaluation of children with strokes outside of the neonatal period, a recommendation supported by the Pediatric Stroke Working Group of the Royal College of Physicians in the United Kingdom.29 It should be emphasized that the high recurrence rates we observed in this study reflect not natural history risk, but rather risk despite current “best medical practice” among a cohort of children with ready access to medical care, many of whom were treated with antithrombotic agents. We need pediatric secondary stroke prevention trials to improve and standardize strategies for reducing recurrent stroke risk in children, particularly among those with abnormal vascular imaging.
This work was supported by an American Heart Association Scientific Development grant and National Institute of Neurological Disorders and Stroke Neurological Sciences Academic Development Award K12 NS01692.
We thank Dr Jean Hayward for thoughtful review of the manuscript.
- Accepted October 31, 2006.
- Address correspondence to Heather J. Fullerton, MD, MAS, Department of Neurology, University of California, Box 0114, 513 Parnassus Ave, S-784, San Francisco, CA 94143-0114. E-mail:
The authors indicated they have no financial relationships relevant to this article to disclose.
- ↵Kurnik K, Kosch A, Strater R, Schobess R, Heller C, Nowak-Gottl U. Recurrent thromboembolism in infants and children suffering from symptomatic neonatal arterial stroke: a prospective follow-up study. Stroke.2003;34 :2887– 2892
- ↵Kirkham FJ, Prengler M, Hewes DK, Ganesan V. Risk factors for arterial ischemic stroke in children. J Child Neurol.2000;15 :299– 307
- ↵Monagle P, Chan A, Massicotte P, Chalmers E, Michelson AD. Antithrombotic therapy in children: the Seventh ACCP Conference on Antithrombotic and Thrombolytic Therapy. Chest.2004;126(3 suppl) :645S– 687S
- ↵Kuhle S, Mitchell L, Andrew M, et al. Urgent clinical challenges in children with ischemic stroke: analysis of 1065 patients from the 1–800-NOCLOTS pediatric stroke telephone consultation service. Stroke.2006;37 :116– 122
- ↵Wu YW, Croen LA, Shah SJ, Newman TB, Najjar DV. Cerebral palsy in a term population: risk factors and neuroimaging findings. Pediatrics.2006;118 :690– 697
- ↵Vittinghoff E, Glidden DV, Shiboski SC, McCulloch CE. Regression Methods in Biostatistics: Linear, Logistic, Survival, and Repeated Measures Models. New York, NY: Springer; 2005
- ↵StataCorp. Stata 9 Reference Manual. College Station, TX: Stata Press; 2005
- ↵Drolet BA, Dohil M, Golomb MR, et al. Early stroke and cerebral vasculopathy in children with facial hemangiomas and PHACE association. Pediatrics.2006;117 :959– 964
- ↵Chabrier S, Rodesch G, Lasjaunias P, Tardieu M, Landrieu P, Sebire G. Transient cerebral arteriopathy: a disorder recognized by serial angiograms in children with stroke. J Child Neurol.1998:13 :27– 32
- ↵Scott RM, Smith JL, Robertson RL, Madsen JR, Soriano SG, Rockoff MA. Long-term outcome in children with moyamoya syndrome after cranial revascularization by pial synangiosis. J Neurosurg.2004;100(2 suppl) :142– 149
- ↵Husson B, Rodesch G, Lasjaunias P, Tardieu M, Sebire G. Magnetic resonance angiography in childhood arterial brain infarcts: a comparative study with contrast angiography. Stroke.2002;33 :1280– 1285
- ↵The Pediatric Stroke Working Group. Stroke in childhood: clinical guidelines for diagnosis, management, and rehabilitation. Available at: www.rcplondon.ac.uk/pubs/books/childstroke. Accessed June 21, 2006
- Copyright © 2007 by the American Academy of Pediatrics