Abstract
BACKGROUND. Ischemic stroke is a known complication of varicella disease. Although there have been case reports of ischemic stroke after varicella vaccination, the existence and magnitude of any vaccine-associated risk has not been determined.
OBJECTIVE. The purpose of this work was to determine whether varicella vaccination is associated with an increased risk of ischemic stroke and encephalitis in children within 12 months after vaccination.
PATIENTS AND METHODS. We conducted a retrospective cohort study based on computerized data from children 11 months through 17 years old enrolled for ≥12 months in the Vaccine Safety DataLink from 1991 through 2004. International Classification of Disease codes identified cases of ischemic stroke (433–436, 437.1, 437.4, 437.6, 437.8–437.9) and encephalitis (052.0, 323.5, 323.8–9). Cox regression was used to model the risk in the 12 months after vaccination relative to all other person-time. Covariates included calendar time, gender, and stroke risk factors (eg, sickle cell disease).
RESULTS. Varicella vaccine was administered to 35.3% of the 3.2 million children in the cohort. There were 203 new inpatient ischemic stroke diagnoses, including 8 that occurred within 12 months after vaccination; there was no temporal clustering. The adjusted stroke hazard ratio was not elevated during any of the time periods in the 12 months after vaccination. Stroke was strongly associated with known risk factors such as sickle cell disease and cardiac disease. None of the 243 encephalitis cases occurred during the first 30 days after vaccination, and there was no association between encephalitis and varicella vaccination at any time in the 12 months after vaccination.
CONCLUSION. Our retrospective cohort study of >3 million children found no association between varicella vaccine and ischemic stroke.
Stroke during childhood is an uncommon and underrecognized source of morbidity and mortality, but it can have profound consequences. Incidence rates for stroke in children beyond the neonatal period estimated from population-based studies vary widely from 1.3 to 13.0 per 100 000 persons per year.1,2 Although most children survive, neurologic or cognitive deficits may persist for ≥50% of subjects.3 Approximately ≥50% of the strokes in children can be classified as ischemic, with the remainder as hemorrhagic.4 The most common risk factors for childhood stroke are cardiac and vascular disease; infection with wild-type varicella zoster virus (VZV) is considered an underappreciated cause of delayed-onset ischemic stroke.5,6 The initial association between VZV and ischemic stroke was established in adults after an episode of herpes zoster ophthalmicus.7 More recently, investigations have established in children the association between primary infection with VZV (ie, chicken pox) and ischemic stroke.6,8–10 Neurologic manifestations of stroke may occur 1 week to 12 months after varicella disease, but signs generally occur within 4 to 12 weeks.6,9
Varicella vaccine contains live attenuated virus and was recommended in 1996 as a routine immunization for young children.11 Case reports since then have suggested a possible association between varicella vaccine and ischemic stroke. One case was based on a report to the US Vaccine Adverse Event Reporting System (VAERS).12 A second report described 2 children with acute-onset hemiparesis shortly after vaccination.13 Although the pathology was consistent with varicella vasculopathy, both children had comorbidities that may have put them at higher risk for stroke (patent foramen ovale and severe iron-deficiency anemia). Taken together with the established association between ischemic stroke and infection with wild-type varicella virus, these reports support an investigation of ischemic stroke as a potential adverse event after vaccination. Therefore, we performed a population-based evaluation of the association between varicella vaccination and ischemic stroke during the 12 months after vaccination among children <18 years of age. In addition, we sought to evaluate the risk of encephalitis after varicella vaccination. Encephalitis is a known complication of varicella disease,14 and a recent study based on VAERS data suggested a temporal association between the vaccine and encephalopathy and meningitis.15
PATIENTS AND METHODS
We conducted a retrospective cohort study based on members of the 8 medical care organizations (MCOs) that participate in the Vaccine Safety DataLink (VSD), a federally funded project designed to evaluate the safety of vaccines.16 Sites participating in VSD contribute computerized immunization, medical encounter (eg, ambulatory and emergency department visits and hospitalizations), and demographic data using a common protocol. The total number of persons of all ages enrolled in any 1 year is ∼8.8 million, and the annual birth cohort is ∼95 000. The study was approved by the institutional review board of each of the participating sites.
The study population was composed of 3.25 million children who were 11 months to 17 years of age during the study period, January 1, 1991, through December 31, 2004. Infants <11 months of age were not studied to avoid including cases of neonatal stroke and because they were unlikely to have been vaccinated with varicella vaccine. Subjects were eligible and could contribute person-time if they had ≥12 months of continuous enrollment (or were enrolled since birth if <12 months) and ≥1 encounter with the MCO while in the age range and study period. Children stopped contributing person-time when they experienced one of the outcomes under study, reached their 18th birthday, disenrolled from the MCO, or the study period ended, whichever came first. Because varicella vaccine was recommended to be given at ≥12 months of age, we excluded persons with diagnoses of infantile cerebral palsy (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] 343.xx), stroke (defined below), or hemiplegia/hemiparesis (ICD-9-CM 342.xx) at ≤11 months of age. We also excluded all of the person-time that occurred after diagnoses of leukemia/lymphoma, HIV/AIDS, primary immune system and bone marrow disorders, leukopenia, myeloproliferative diseases, and other syndromes associated with immunodeficiency, because these illnesses may be independently associated with stroke-like syndromes. Diagnosis codes were from the ICD-9-CM.
The primary outcome was ischemic stroke identified using primary and secondary coded diagnoses assigned in the inpatient and/or outpatient settings (ICD-9-CM 433–436, 437.1, 437.4, 437.6, and 437.8–437.9). Similarly, we examined diagnoses of encephalitis as a secondary outcome (ICD-9-CM 052.0, 323.5, 323.8, and 323.9). Because stroke and encephalitis are serious illnesses, we focused our investigation on outcomes that resulted in hospitalizations (ie, diagnoses made in the inpatient setting), but we also examined outpatient diagnoses as a secondary outcome. We based our analyses on coded diagnoses; they were not confirmed by review of medical charts, and there was no attempt to standardize the diagnostic evaluation. The risk window was defined as the 12-month period after receipt of varicella vaccine; all of the other person-time was considered unexposed. The 12-month risk period was selected because reports have suggested that the incidence of stroke rarely exceeds 1 year after VZV infection.6,9,10
ICD-9-CM diagnosis codes were used to identify conditions that may increase the risk of stroke and included cardiac disease (eg, cardiomyopathy and septal defects), rheumatic heart disease and endocarditis, cardiovascular conditions (eg, hypertensive disease), sickle cell disease, genetic conditions predisposing to vasculopathy (eg, Down syndrome), coagulation abnormalities, and diseases leading to a hypercoagulable state (eg, systemic lupus erythematosus). In the analysis, these factors were classified as present on the first diagnosis found in the electronic medical chart. Race was not reliably available and was not examined.
Cox regression was used to evaluate the association between the exposures (varicella vaccination) and outcomes (ischemic stroke and encephalitis). The counting process formulation was used to account for time-dependent covariates, left truncation, and discontinuities in at-risk time.17 The time origin in the regression modeling was date of birth. All of the models were stratified by VSD site and included the following independent variables: exposure (<1, 1–2, 3–5, 6–8, and 9–11 months after an exposure date, as well as all other at-risk times as the referent, unexposed category), gender, calendar time (1991–1995 [referent category], 1996–1998, 1999–2001, and 2002–2004), and dichotomous variables corresponding with the presence of stroke risk factors. Censoring dates included diagnosis dates for conditions likely to cause immune suppression or deficiency (as described above), date of death, and the date on which the person was no longer within the age range and observation period of the study. Follow-up time was terminated on first occurrence of the outcome.
Based on preliminary estimates of children 1 to 6 years old who were assigned ischemic stroke diagnosis codes, we estimated the power to be 80% to detect a relative risk of 1.7 for stroke during a 9-month postvaccination period and 96% to detect a relative risk of 2.0.
RESULTS
Approximately 1.14 million children (35.3%) of the 3.25 million children in the study cohort had ≥1 varicella vaccination during the study period (Table 1). Excluded from the analysis were 10 105 children with infantile cerebral palsy (n = 8712) and diagnoses before 11 months old of stroke (n = 189) or hemiplegia/hemiparesis (n = 1204). The total number of person-years of observation was 17.2 million, of which 54% were contributed by children ≤9 years old; 64% of person-time occurred in the latter half of the study period (1998–2004). Of those vaccinated, 79.6% were <5 years old at the time of vaccination, and 58.3% were vaccinated between the ages of 1 and 2 years. Because the vaccine was not widely distributed until the late 1990s, unvaccinated children were, on average, substantially older than vaccinated children on entering the study cohort (7.9 vs 1.9 years; P < .0001). Boys constituted slightly more than half of the cohort in both the vaccinated and nonvaccinated groups.
Characteristics of the Varicella Vaccine-Ischemic Stroke Study Cohort According to Varicella Vaccination Status
Ischemic Stroke
There were 203 children with ischemic stroke diagnoses assigned in the inpatient setting, resulting in an ischemic stroke incidence rate of 1.2 per 100 000 person-years (95% confidence interval [CI]: 1.0–1.4). Although rare in both groups, stroke was more common among children who did not have a varicella vaccination compared with those who did, even after adjustment for age (0.008% vs 0.003%; P < .0001) (Table 1). Fifty-seven percent of all strokes were among boys (P = .08); the age at onset was similar for girls and boys (median age: 13 vs 12 years; P < .91). Of the 203 children with ischemic stroke diagnoses, 39 (19.2%) were vaccinated on or before their diagnosis; the median interval between vaccination and stroke onset was ∼4 years. Of the 8 cases that occurred within 12 months after vaccination, 1 subject was diagnosed on the day of vaccination, and only 1 other case occurred within the first 3 months after vaccination. Overall, there was no evidence of temporal clustering after varicella vaccination. The most commonly assigned stroke diagnosis code was for occlusion of the cerebral arteries (ICD-9-CM 434.xx) (Table 2).
Frequency of Ischemic Stroke or Encephalitis Diagnosis Codes Assigned to Children in the Hospital Setting
The time from vaccination to ischemic stroke was evaluated using a Cox regression model. The adjusted hazard ratio (HR) for stroke was not significantly elevated at any time in the 12 months after vaccination (Table 3). The HR was 1.1 (95% CI: 0.1–9.2) in the first 30 days after vaccination and 0.7 (95% CI: 0.1–5.7) in the following 60 days. We repeated the analysis using only the exposed and unexposed person-time from children who had been administered the varicella vaccine; the results were essentially the same (data not shown).
Number of Children With Ischemic Stroke and Adjusted HRs for Ischemic Stroke According to Time Since Varicella Vaccination
The proportion of children with ≥1 risk factor for stroke was not substantially different between the vaccinated and nonvaccinated groups (Table 1). Among the 203 subjects with ischemic stroke, risk factors were identified in 87 (42.9%; Table 4). For both groups, ischemic stroke was strongly associated with the presence of ≥1 risk factor, as well as most of the individual risk factors, including cardiac disease, systemic lupus erythematosus, and sickle cell disease. For example, the HR was 170 (95% CI: 97–299) for sickle cell disease and 20 (95% CI: 13–31) for cardiac conditions, such as cardiomyopathy and congenital valvular disease (Table 3). Girls were somewhat less likely to have a stroke diagnosis than boys (HR: 0.7 [95% CI: 0.5–1.0]).
Number and Percentage of Children With Risk Factors for Ischemic Stroke According to Varicella Vaccination and Stroke Status
We also evaluated the risk associated with vaccination when ischemic stroke was diagnosed in either the outpatient or inpatient settings. Of the 833 cases, 42 occurred in the 12 months after vaccination. The adjusted HR in the first 90 days after vaccination was 1.2 (95% CI: 0.7–2.3) and ranged from 1.2 to 0.9 to 1.1 in the 3 ensuing 3-month periods, respectively; each of the 95% CIs included the null. The HRs associated with the various risk factors mirrored those when considering only inpatient cases, although the HR magnitude was generally reduced. For example, the HRs for sickle cell disease and cardiac conditions were 55.0 (95% CI: 35.2–85.9) and 8.0 (95% CI: 6.0–10.7), respectively.
Encephalitis
A total of 243 children were diagnosed with encephalitis in the inpatient setting (Table 2). Forty-seven children (19.3%) were vaccinated for varicella on or before the date of their encephalitis diagnosis; the median interval between vaccination and disease onset was 2.7 years. Of the 47 cases, 11 (23.4%) occurred in the 12 months after vaccination. None of the 11 case subjects were diagnosed in the first 30 days, and only 1 was diagnosed within the first 90 days after vaccination. Cox regression was used to model the time from vaccination to encephalitis with the same stratification and covariates that were used in the stroke model. The adjusted HR for encephalitis was not significantly elevated at any time in the 12 months after vaccination (data not shown). For example, the risk of encephalitis among vaccinated compared with nonvaccinated subjects was 0.7 (95% CI: 0.1–5.2) for the 30 to 90 days after varicella vaccination.
DISCUSSION
Investigations have established the relationship between primary VZV infection and ischemic stroke in children,6,18 but we are aware of no investigation that has systematically examined the association between varicella vaccination and ischemic stroke in a large, well-defined population. Our retrospective cohort study based on administrative data from >3 million children found no association between the vaccine and ischemic stroke. We observed relatively few cases of ischemic stroke in the 12 months after varicella vaccination, and there was no evidence of temporal clustering. Multivariable models revealed strong associations between ischemic stroke and established risk factors but no increase in the risk of ischemic stroke at any time in the 12 months after varicella vaccination.
The rate of ischemic stroke in our population (1.2 per 100 000 person-years) was nearly identical to that in 2 large, population-based investigations. Fullerton et al4 studied 1167 hospital discharges from 1991 through 2000 in California and reported an annual incidence rate of 1.2 per 100 000 children. Broderick et al19 reviewed all of the records for children with stroke in the Cincinnati area in 1988–1989 and reported an incidence rate of 1.2 per 100 000 (95% CI: 0.3–2.0). However, both higher (2.6 per 100 000 [95% CI: 2.5–2.9])20 and lower rates (0.6 per 100 000 [95% CI: 0.4–1.3])1 have been reported. Caution is needed when comparing rates of stroke incidence for several reasons, including differences in study design and population, secular trends in stroke etiologies and diagnostics, and differences in case definition. For example, our study population excluded infants ≤11 months old, a group that suffers a larger proportion of stroke compared with older children.21 However, the lower age range for our study cohort was determined by official recommendations for varicella vaccination (12–15 months).11
A large number of conditions predispose children to ischemic stroke, but the underlying cause may be unknown in ≤50% of children, and a substantial proportion may be detected only after the stroke has been diagnosed.10,22 In our study, 41% of the case subjects had been assigned the diagnostic code of a risk factor before the stroke diagnosis, which is somewhat lower than the 50% reported by deVeber.22 Cardiac and vascular conditions were the most frequently identified risk factors among children with ischemic stroke in our population (19% and 16%, respectively), in accord with other reports.3,4,23
Encephalitis is a rare complication after varicella disease,14 and a possible association with varicella vaccine has been reported recently. Using data from VAERS, Chaves et al15 reported a temporal association between varicella vaccine and cases of encephalopathy. In addition, the vaccine strain of VZV was isolated from 2 meningitis cases that occurred subsequent to the development of herpes zoster. However, we found no association between varicella vaccination and encephalitis and only 1 case of encephalitis within 90 days of vaccination; we did not examine meningitis for this study.
Our study is subject to a number of limitations. We identified case subjects by searching the computerized data of the MCOs in VSD for coded diagnoses of ischemic stroke. We focused on first-ever inpatient as opposed to outpatient diagnoses, because most children with acute disease would be admitted to the hospital, and such diagnoses should have a higher level of validity. Because we observed no suggestion of an association between varicella vaccination and ischemic stroke, these diagnoses were not verified by chart review. Studies of hospitalized adults, which account for most of the studies examining the validity of stroke codes, have typically found moderately good sensitivity and positive predictive values for the standard stroke codes.24,25 However, comparable studies among children are scarce. In addition, manifestations of ischemic stroke in children can be more subtle, and their symptoms are more apt to be attributed to other stroke-like conditions (eg, migraine), which may lead to misdiagnoses or delays in diagnosis.22,26 A recent study of ischemic strokes in children from a single hospital found that the code-specific accuracy of standard stroke codes ranged from 37% to 88%, although they included in their computations both inpatients and outpatients.27 The validity of stroke and encephalitis ICD-9-CM codes has not been formally evaluated in VSD. However, VSD studies of other diseases identified in the inpatient setting (eg, intussusception) have reported relatively high predictive values.28 It is reassuring that our rate of ischemic stroke is comparable to the lower range of rates found in other population-based studies.4,19 In addition, our male/female stroke ratio agrees with other investigations,4,22,23 as does the prevalence and composition of pre-existing risk factors.22 Although hemorrhagic strokes account for ∼50% of all childhood strokes,4,19 we did not include them in our case ascertainment, because only ischemic stroke has been recognized as a complication of infection with wild-type VZV.6,22 It should be noted, however, that a single case of intracranial hemorrhage 2 months after varicella disease was reported recently.29
Selecting only hospitalized case subjects could introduce survivor bias by excluding children who rapidly succumbed to their disease. However, this is likely to be a very small fraction of all stroke case subjects.3 At the opposite end of the disease spectrum, bias could also be introduced by the exclusion of children whose disease manifestations were less severe, resulting in a diagnosis during an office visit and no subsequent hospitalization. However, incorporation of outpatient diagnoses of ischemic stroke into our outcome variable did not alter our findings of no association with vaccination.
Our study captured ischemic stroke cases over a 14-year period, during which time diagnostic techniques have improved in their sensitivity and specificity.30 This is unlikely to have a substantial effect on our results, because 66% of our cases occurred in the second half of the study period, and our regression models adjusted for calendar year. Our computerized vaccination information is likely to be of high quality, as demonstrated in a previous investigation.31 Finally, our findings may not be generalizable to the population as a whole, because MCO members may be healthier than others.32
CONCLUSIONS
Although diagnosis of stroke in the United States may be increasing, in part, because of advances in imaging techniques and improved survival of children with conditions that may predispose to stroke,3,30 varicella-associated stroke should decrease in parallel with the decreasing incidence of varicella disease. Complementing 2 recent reviews that found serious adverse events to be rare after varicella vaccination,15,33 this study offers reassurance that the rare complication of stroke seen after varicella infection is not seen after vaccination with the live attenuated vaccine.
Acknowledgments
Financial support for this study was provided in full by the Centers for Disease Control and Prevention (contract 200-2002-00732), through America's Health Insurance Plans.
We thank Dr Maitreyi Mazumdar for help in specifying codes for ischemic stroke and its risk factors. We also thank the following members of the VSD team for their efforts in data management, project administration, and programming: Nita Herr and Deb Kempf (Marshfield Clinic Research Foundation); Amy Butani and Leslie Kuckler (Health Partners); Patti Benson and Darren Malais (Group Health Cooperative); Rich Fox and Charlene Gay (Harvard Pilgrim Health Care); Marcia Cunningham, Jo Ann Shoup, and Kristi Yamasaki (Kaiser Permanente Colorado); Loie Drew, Jill Mesa, and Karen Riedlinger (Kaiser Permanente Northwest); Julianne Gee and Eric Weintraub (Centers for Disease Control and Prevention); and the staff of America's Health Insurance Plans.
Footnotes
- Accepted October 10, 2008.
- Address correspondence to James G. Donahue, DVM, PhD, Marshfield Clinic Research Foundation, Epidemiology Research Center, ML-2, 1000 N Oak Ave, Marshfield, WI 54449. E-mail: donahue.james{at}mcrf.mfldclin.edu
Financial Disclosure: Dr Marcy reports serving as a consultant to Merck Pharmaceuticals, Dr Baxter and Mr Lewis report receiving research support from Merck, and Dr Klein reports receiving research support from Merck, Novartis, GlaxoSmithKline, Sanofi-Pasteur, and MedImmune; the other authors report no potential conflicts of interest relevant to this research.
The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.
What's Known on This Subject
Ischemic stroke during childhood is an uncommon but serious event. Childhood stroke has many causes and is a known complication after varicella disease. However, the association between stroke and the live virus varicella vaccine has not been reported.
What This Study Adds
This study examined the 12-month period after varicella vaccination and found no association with ischemic stroke among the 3 million children in the VSD. In addition, we found no association between varicella vaccination and encephalitis.
REFERENCES
- Copyright © 2009 by the American Academy of Pediatrics
In this issue
