- Top of page
- What this paper adds
Aim To examine maternal hypertension, diabetes, and intrapartum fever as potential risk factors for ischemic stroke in infants.
Method We conducted a retrospective cohort study of 226 117 children born from January 2000 to December 2007 who were enrolled in the South Carolina Medicaid program. We linked maternal and child Medicaid billing records and birth certificate data. Children with ischemic stroke were identified based on the International Classification of Diseases, Ninth Revision (ICD-9), code 434 in the child’s billing data. Independent variables and covariates were identified using ICD-9 codes and birth certificate data. We modeled the odds of ischemic stroke diagnosis in infants, either before 30 days of life or before 365 days.
Results Forty-three children were diagnosed with ischemic stroke before 30 days and 161 before 365 days. Maternal hypertension (odds ratio 2.31 before 30d) and intrapartum fever (odds ratio 3.36 <30d) were significantly associated with odds of ischemic stroke before 30 days and before 365 days; maternal diabetes was not.
Interpretation Maternal hypertension and intrapartum fever appear to be risk factors for ischemic stroke in infants. Additional research is needed to determine the mechanism(s) underlying these associations and to develop effective preventive methods for high-risk infants.
Ischemic stroke is a rare condition in children, occurring most frequently in the neonatal period. Approximately two to four children per 10 000 live births experience ischemic stroke in the first 28 days of life.1 Approximately 60% of infants with neonatal stroke present immediately, most frequently with neonatal seizures.2 In the remaining 40%, stroke is recognized later in childhood during evaluation for abnormal neurological or cognitive development.2 Long-term sequelae of neonatal ischemic stroke include epilepsy, cerebral palsy, intellectual disability, hemiparesis, and other neurological disabilities.3 Risk of perinatal ischemic stroke is higher in infants with conditions that predispose to thrombosis; such conditions are genetic (such as protein S or C deficiency), infectious (such as meningitis or sepsis), and cardiovascular (such as congenital heart defects and cardiac surgery), as well as others.4 Lee et al.5 reported that significant prenatal and perinatal risk factors for perinatal arterial ischemic stroke included a history of infertility, oligohydramnios, pre-eclampsia, prolonged rupture of membranes, umbilical cord abnormality, chorioamnionitis, and primiparity. Cheong and Cowan1 identified five studies that reported pre-eclampsia was a significant risk factor.
Hypertension during pregnancy can be described as chronic/non-gestational, pregnancy-induced or gestational (onset during pregnancy), pre-eclampsia (with proteinuria), or eclampsia (with proteinuria and seizures).6,7 In the USA, pre-pregnancy hypertension is present in approximately 2% of pregnancies, gestational hypertension without pre-eclampsia in 2 to 3%, and pre-eclampsia or eclampsia in approximately 3%.6 Pregnancy-induced hypertension and pre-eclampsia/eclampsia are thought to be initiated by problems with placental development that ultimately lead to an immune response and other physiological changes that produce hypertension and other manifestations of disease.8 Maternal hypertension is a known cause of preterm delivery,9 intrauterine growth restriction,10 and placental abruption.11 We have previously reported associations between maternal pre-eclampsia and intellectual disability, epilepsy, and cerebral palsy.12–14
Between 4% and 14% of pregnant females in the USA experience gestational diabetes mellitus,15 not including females with pre-existing diabetes. Maternal diabetes is a risk factor for preterm labor, maternal hypertension, pregnancy loss, macrosomia, and neonatal hypoglycemia.16 Edmonds et al.17 described a series of neonates with ischemic stroke, and reported that 4 out of 10 had mothers with gestational diabetes; however, we are not aware of adequately powered studies reporting such an association.
We conducted a retrospective cohort study of births reimbursed by the South Carolina Medicaid program, which is a state run health insurance program for low-income individuals and pays for approximately half of all births in South Carolina, to test the hypotheses that maternal hypertension, and diabetes, and intrapartum fever are associated with increased risk of neonatal ischemic stroke. We chose to focus on these maternal characteristics because they are relatively common pregnancy complications, because there is insufficient information about whether they are associated with increased risk, and because identification of risk factors during pregnancy may have the potential to lead to the development of effective preventive interventions for high-risk pregnancies.
- Top of page
- What this paper adds
The study protocol was approved by the Institutional Review Board at the University of South Carolina. We obtained linked maternal–child data for all births reimbursed by the South Carolina Medicaid program from January 1, 2000 to December 31, 2007. After excluding the 6581 non-singleton births, 226 117 observations were used in analyses. Independent variables and covariates were identified in both Medicaid billing data and linked birth-certificate data. For example, females with hypertension were identified based on a diagnosis of essential hypertension, secondary hypertension, hypertensive heart or kidney disease, pregnancy-induced hypertension, or pre-eclampsia/eclampsia in the Medicaid data, or identification with pre-pregnancy hypertension, pregnancy induced hypertension, or eclampsia in the birth certificate data. Females with diabetes were identified based on a diagnosis with type 1 or type 2 diabetes, gestational diabetes, or unspecified diabetes complicating pregnancy in the billing data, or with chronic or gestational diabetes on birth certificates. Intrapartum fever and maternal tobacco use were identified using birth certificate data.
The outcome studied was diagnosis with ischemic stroke (‘occlusion of cerebral arteries,’ International Classification of Diseases, Ninth Revision [ICD-9] code 434) in the Medicaid billing data, with initial diagnosis up to 30 days after birth. Though the neonatal period is typically defined as lasting through day 28, we included the additional day because of the possibility that a stroke could occur on day 28 but not be billed until the following day. We also analysed all ischemic stroke diagnosed before 365 days of age, because most of these were probably manifestations of neonatal strokes.2 The complete list of variables is shown in Table I, which includes a description of each variable as well as the data source and ICD-9 codes (if applicable).
Table I. Summary of variables associated with neonatal ischemic stroke
| Any hypertension||Chronic/non-gestational hypertension||MB||401–405, 642.0–642.2|
|Gestational hypertension (pregnancy induced hypertension, pre-eclampsia, eclampsia)||MB and BC||642.3–642.6|
|Pre-eclampsia/eclampsia superimposed on chronic hypertension||MB||642.7|
| Any diabetes||Chronic diabetes (type 1 or 2)||MB and BC||248, 250, 648.0|
|Gestational diabetes||MB and BC||648.8|
| Intrapartum Fever||Temperature ≥38°C (2000–2003); temperature >38°C or clinical chorioamnionitis diagnosed during labor (2004–2007)||BC||NA|
| Tobacco use||Any use in pregnancy||BC||NA|
| Maternal age||Years||BC||NA|
| Maternal race||White, black, other||BC||NA|
| Maternal ethnicity||Hispanic or non-Hispanic||BC||NA|
| Placental abruption||Any placental abruption||MB and BC||641.2|
| Gestational age||Completed weeks||BC||NA|
| Child’s sex||Male or female||BC||NA|
| Prothrombotic conditions||Sickle cell trait||MB||282.5|
|Sickle cell disease||MB||282.6|
|Primary hypercoagulable state||MB||289.8|
| Meningitis||Bacterial, viral, or unspecified meningitis||MB||320–322, 047–049, 036|
| Encephalitis||Encephalitis from any cause||MB||046, 049, 052.0, 055.0, 056.0, 058.2, 062–064, 072.2, 130.0, 139.0, 323|
| Congenital infections||Congenital: rubella pneumonitis, cytomegalic inclusion disease, herpes simplex, listeriosis, malaria, toxoplasmosis, tuberculosis||MB||771.0–771.2|
| Neonatal infections||Septicemia of newborn, bacteremia of newborn, other infections specific to the perinatal period||MB||771.81, 771.83, 771.89|
| Birth asphyxia||Severe birth asphyxia, mild or moderate birth asphyxia||MB and BC||768.5, 768.6, 768.0|
| Birth trauma||Any diagnosis of birth trauma, brain or other||MB and BC||767|
| Neurodevelopmental disability||Hemiplegia/hemiparesis||MB||342|
|Late effects of stroke||MB||438|
| Ischemic stroke||Occlusion of cerebral arteries, yes or no||MB||434|
We modeled the odds of ischemic stroke using logistic regression. We created separate models for ischemic stroke diagnosed before 30 days, then expanded the period to include all ischemic stroke diagnosed before 365 days. In each model, we excluded children diagnosed with ischemic stroke after the end of the period of interest (≥30d or ≥365d). Because it is possible for the same woman to have multiple births in the study period, there is a potential for lack of independence among observations. One way to deal with clustered observations is to estimate generalized estimating equation models,18 though if the within-cluster dependency is very weak, ordinary logistic regression models (for independent data) are appropriate.18–20 When we fitted the generalized estimating equations models with an exchangeable correlation structure, the working correlation was estimated to be less than 0.01. Given that dependencies among children born to the same mother were very low, we estimated the models using ordinary logistic regression (PROC LOGISTIC in SAS version 9.2; SAS Institute, Cary, NC, USA).
For some variables, we encountered the problem of quasi-complete separation, which can occur when several risk factors nearly perfectly predict the response and results in maximum likelihood estimates that are non-unique and infinite. To address this problem, we used Firth’s method,21 which is considered the best solution to separation in logistic regression.22 Firth’s method estimates the regression coefficients by maximizing a penalized likelihood function. It provides finite parameter estimates for risk factors that cause separation and further reduces bias in the estimation.
To avoid overtaxing the models with unnecessary covariates, we performed variable selection based on Akaike information criterion. For each model, the set of covariates producing the smallest Akaike information criterion value is chosen. Maternal hypertension, diabetes, and fever were retained in all the models, because they are the primary independent variables for the study.
We used two different case definitions for ischemic stroke for each of the periods for stroke diagnosis. The first definition included all children with at least one diagnosis of ischemic stroke in the period of interest. The second case definition represented ‘confirmed’ cases of ischemic stroke and required at least one diagnosis of ischemic stroke within the period of interest plus at least one of either (1) another diagnosis of ischemic stroke either during the same period or later in childhood, or (2) diagnosis of one of the neurodevelopmental disabilities (hemiplegia/hemiparesis, ‘late effects of stroke,’ cerebral palsy, intellectual disability, or epilepsy) at some time in childhood.
- Top of page
- What this paper adds
Forty-three children had at least one diagnosis of ischemic intracerebral stroke before 30 days of age; 37 of these children had confirmed ischemic stroke. An additional 118 children were diagnosed with ischemic stroke after 29 days but before 365 days, 99 of whom had confirmed ischemic stroke. Characteristics of children without ischemic stroke, those with ischemic stroke diagnosed before 30 days, and those diagnosed before 365 days are compared in Table II. Owing to missing values, sums of frequency counts may not equal the corresponding column totals. Maternal hypertension was more common for infants with ischemic stroke before 30 days and before 365 days. Intrapartum fever was significantly more common in mothers of infants with ischemic stroke before 365 days but not before 30 days (though the proportions with maternal fever were very similar). There was not a significant difference in the prevalence of maternal diabetes, for infants with ischemic stroke either before 30 days or before 365 days.
Table II. Cohort characteristics
| ||Without ischemic stroke (n=225 913)||Ischemic stroke before 30d (n=43)||Ischemic stroke before 365d (n=161)|
|Variable|| n (or mean)a||% (or SD)|| n (or mean)||% (or SD)|| pb || n (or mean)||% (or SD)|| p |
|Maternal age (y)||23.7||5.4||24.4||5.5||0.373||23.8||5.4||0.864|
|Gestational age (wks)||38.4||2.2||35.8||4.2||0.002||35.3||5.1||<0.001|
|Missing 565||Missing 1|
|Preterm||Term 199 934||88.7||Term 25||58.1||<0.001||Term 92||57.5||<0.001|
|MPT 19 754||8.8||MPT 10||23.3||MPT 27||16.9|
|VPT 5660||2.5||VPT 8||18.6||VPT 41||25.6|
|Missing 565||Missing 1|
|Maternal race||White 115 547d||51.2||White 24||55.8||0.607||White 79||49.1||0.443|
|Black 106 287||47.1||Black 19||44.2||Black 81||50.3|
|Other 3888||1.7||Other 0||0.0||Other 1||0.6|
|Maternal Hispanic ethnicity||No 202 758||89.9||No 39||90.7||1.000e||No 146||90.7||0.749|
|Yes 22 717||10.1||Yes 4||9.3||Yes 15||9.3|
|Maternal education ≥12y||Yes 142 677||63.5||Yes 27||62.8||0.928||Yes 99||62.7||0.836|
|No 82 175||36.6||No 16||37.2||No 59||37.3|
|Missing 1061||Missing 3|
|Child sex||Male 115 469||51.1||Male 24||55.8||0.537||Male 95||59.0||0.045|
|Female 110 443||48.9||Female 19||44.2||Female 66||41.0|
|Tobacco use during pregnancy||No 184 006||81.5||No 32||74.4||0.229||No 128||79.5||0.507|
|Yes 41 672||18.5||Yes 11||25.6||Yes 33||20.5|
|Maternal fever at delivery||No 223 205||98.8||No 41||95.4||0.094e||No 153||95.0||<0.001e|
|Yes 2706||1.2||Yes 2||4.6||Yes 8||5.0|
|Maternal diabetes||No 200 858||88.9||No 41||95.4||0.228e||No 146||90.7||0.474|
|Yes 25 055||11.1||Yes 2||4.7||Yes 15||9.3|
|Maternal hypertension||No 194 525||86.1||No 29||67.4||0.004||No 116||72.1||<0.001|
|Yes 31 386||13.9||Yes 14||32.6||Yes 45||27.9|
|Child birth trauma||No 219 344||97.1||No 34||79.1||<0.001e||No 131||81.4||<0.001e|
|Yes 6569||2.9||Yes 9||20.9||Yes 30||18.6|
|Child birth asphyxia||No 225 678||99.9||No 40||93.0||<0.001e||No 152||94.4||<0.001e|
|Yes 235||0.1||Yes 3||7.0||Yes 9||5.6|
|Child thalassemia||No 225 688||99.9||No 43||100.0||1.000e||No 161||100.0||1.000e|
|Yes 225||0.1||Yes 0||0.0||Yes 0||0.0|
|Child sickle cell disease||No 224 835||99.5||No 40||93.0||0.001e||No 158||98.1||0.043e|
|Yes 1078||0.5||Yes 3||7.0||Yes 3||1.9|
|Child sickle cell trait||No 224 421||99.3||No 41||95.4||0.033e||No 159||98.8||0.288e|
|Yes 1492||0.7||Yes 2||4.6||Yes 2||1.2|
|Child thrombophilia||No 225 883||99.99||No 40||93.0||<0.001e||No 156||96.9||<0.001e|
|Yes 30||0.01||Yes 3||7.0||Yes 5||3.1|
|Child congenital infection||No 225 593||99.9||No 42||97.7||0.059e||No 156||96.9||<0.001e|
|Yes 320||0.1||Yes 1||2.3||Yes 5||3.1|
|Child neonatal infection||No 220 508||97.6||No 27||62.8||<0.001e||No 108||67.1||<0.001e|
|Yes 5405||2.4||Yes 16||37.2||Yes 53||32.9|
|Child meningitis||No 224 494||99.4||No 39||90.7||<0.001e||No 140||87.0||<0.001e|
|Yes 1419||0.6||Yes 4||9.3||Yes 21||13.0|
|Child encephalitis||No 218 715||96.8||No 43||100.0||0.648e||No 146||90.7||<0.001|
|Yes 7198||3.2||Yes 0||0.0||Yes 15||9.3|
|Placental abruption||No 222 535||98.5||No 41||95.4||0.135e||No 147||91.3||<0.001e|
|Yes 3378||1.5||Yes 2||4.6||Yes 14||8.7|
Infants with ischemic stroke before 30 days had lower gestational ages and were more likely to have experienced birth trauma or birth asphyxia, have sickle disease or sickle cell trait, have thrombophilia, have neonatal infection, and to have meningitis. Expanding the timing of initial diagnosis with ischemic stroke, affected infants were also more likely than unaffected infants to be male and to have congenital infection, encephalitis, and maternal placental abruption, but were not more likely to have sickle cell trait.
We estimated models for any and confirmed ischemic stroke, before 30 days and before 365 days, controlling for covariates. As described previously, variables were selected based on Akaike information criterion, and Firth’s method was used whenever quasi-separation occurred. The results are presented in Tables III and IV. In modeling the outcome of any ischemic stroke before 30 days, maternal ethnicity, child encephalitis, placental abruption, and child thalassemia were not selected. For the model of any ischemic stroke before 365 days, child thalassemia was not selected.
Table III. Adjusted odds ratios (OR) for ischemic stroke before 30 daysa
|Variable||Any ischemic stroke||Confirmed ischemic stroke|
|OR (95% CI)|| p ||OR (95% CI)|| p |
|Maternal race (0, white; 1, black; 2, other)||0.76 (0.41–1.40)b||0.378||Not in model||—|
|Maternal ethnicity (Hispanic vs non-Hispanic)|| Not in model||—||Not in model||—|
|Maternal education (<12y vs ≥12y)||1.06 (0.58–1.93)||0.858||Not in model||—|
|Maternal age||1.03 (0.98–1.08)||0.252||1.03 (0.97–1.08)||0.336|
|Child sex (female vs male)||0.90 (0.51–1.58)||0.714||0.70 (0.38–1.31)||0.269|
|Tobacco use during pregnancy (yes vs no)||1.52 (0.78–2.95)||0.220||1.61 (0.80–3.24)||0.186|
|Maternal fever at delivery|| 3.36 (1.01–11.19)|| 0.048 || 4.02 (1.18–13.67)|| 0.026 |
|Maternal diabetes||0.35 (0.11–1.16)||0.085||0.40 (0.12–1.35)||0.139|
|Maternal hypertension|| 2.31 (1.26–4.24)|| 0.007 || 2.75 (1.45–5.24)|| 0.002 |
|Child birth trauma|| 7.06 (3.48–14.30)|| <0.001 || 8.68 (4.21–17.91)|| <0.001 |
|Child birth asphyxia|| 17.15 (5.42–54.23)|| <0.001 || 22.22 (6.97–70.88)|| <0.001 |
|Child sickle cell disease|| 10.40 (3.07–35.28)|| 0.002 || 11.39 (3.29–39.42)|| 0.001 |
|Child sickle cell trait|| 6.92 (1.71–28.04)|| 0.007 || 6.89 (1.69–28.07)|| 0.007 |
|Child thrombophilia|| 367.95 (99.72–1000+)|| <0.001 || 413.17 (111.10–1000+)|| <0.001 |
|Child congenital infection|| 10.47 (2.17–50.67)|| 0.004 || 14.17 (2.83–70.99)|| 0.001 |
|Child neonatal infection|| 11.50 (5.64–23.44)|| <0.001 || 9.53 (4.39–20.69)|| <0.001 |
|Child meningitis||2.69 (0.49–14.72)||0.253||Not in model||—|
|Child encephalitis||Not in model||—||Not in model||—|
|Gestational age (wks)|| 0.92 (0.86–0.99)|| 0.029 ||0.95 (0.87–1.03)||0.176|
|Placental abruption||Not in model||—||Not in model||—|
|Child thalassemia||Not in model||—||Not in model||—|
Table IV. Adjusted odds ratios (OR) for ischemic stroke before 365 daysa
|Variable||Any ischemic stroke||Confirmed ischemic stroke|
|OR (95% CI)|| p ||OR (95% CI)|| p |
|Maternal race (0, white; 1, black; 2, other)||0.91 (0.64–1.30)b||0.613||0.92 (0.63–1.35)b||0.667|
|0.39 (0.06–2.49)c||0.318||0.44 (0.07–2.89)c||0.395|
|Maternal ethnicity (Hispanic vs non-Hispanic)||0.96 (0.52–1.78)||0.896||1.03 (0.54–1.96)||0.934|
|Maternal education (<12y vs ≥12)||1.04 (0.73–1.47)||0.844||1.12 (0.77–1.62)||0.564|
|Maternal age||1.00 (0.98–1.04)||0.782||1.00 (0.97–1.03)||0.971|
|Child sex (female vs male)||0.81 (0.59–1.11)||0.183|| 0.68 (0.48–0.96)|| 0.030 |
|Tobacco use during pregnancy (yes vs no)||1.13 (0.75–1.70)||0.570||1.03 (0.65–1.63)||0.891|
|Maternal fever at delivery|| 3.16 (1.57–6.37)|| 0.001 || 3.04 (1.39–6.65)|| 0.006 |
|Maternal diabetes||0.74 (0.44–1.26)||0.270||0.91 (0.53–1.55)||0.731|
|Maternal hypertension|| 1.93 (1.36–2.74)|| 0.003 || 1.93 (1.32–2.83)|| 0.007 |
|Child birth trauma|| 5.99 (3.95–9.09)|| <0.001 || 5.97 (3.82–9.33)|| <0.001 |
|Child birth asphyxia|| 11.42 (5.25–24.86)|| <0.001 || 11.79 (5.18–26.85)|| <0.001 |
|Child sickle cell disease|| 3.58 (1.17–10.98)|| 0.026 || 4.04 (1.31–12.47)|| 0.015 |
|Child sickle cell trait||2.45 (0.70–8.65)||0.163||2.75 (0.78–9.71)||0.116|
|Child thrombophilia|| 157.99 (54.20–460.56)|| <0.001 || 182.64 (62.70–532.05)|| <0.001 |
|Child congenital infection|| 5.39 (1.90–15.33)|| 0.002 ||3.36 (0.81–13.90)||0.095|
|Child neonatal infection|| 6.06 (3.93–9.33)|| <0.001 || 5.26 (3.26–8.50)|| <0.001 |
|Child meningitis|| 6.05 (3.31–11.06)|| <0.001 || 6.19 (3.18–12.05)|| <0.001 |
|Child encephalitis|| 3.99 (1.67–9.56)|| 0.002 || 3.37 (1.19–9.52)|| 0.022 |
|Gestational age (wks)|| 0.88 (0.85–0.92)|| <0.001 || 0.89 (0.85–0.93)|| <0.001 |
|Placental abruption||1.64 (0.91, 2.97)||0.100||1.80 (0.95, 3.39)||0.071|
|Child thalassemia||Not in model||—||Not in model||—|
Maternal hypertension was significantly associated with any ischemic stroke diagnosed before 30 days (odds ratio [OR] 2.31, 95% confidence interval [CI] 1.26–4.24) and before 365 days (OR 1.93, 95% CI 1.36–2.74). Intrapartum fever was also significantly associated with any ischemic stroke both before 30 days (OR 3.36, 95% CI 1.01–11.19) and before 365 days (OR 3.16, 95% CI 1.57–6.37). Maternal diabetes was not significant in either model. Gestational age, birth trauma, birth asphyxia, child thrombophilia, sickle cell disease, congenital infection, and neonatal infection were significantly associated with ischemic stroke in both models. Child sickle-cell trait was a significant risk factor for ischemic stroke before 30 days but not when including strokes up to 364 days. Meningitis was a risk factor for ischemic stroke before 365 days but not before 30 days.
We repeated the model building for confirmed ischemic stroke. Maternal hypertension and intrapartum fever were significantly associated with odds of ischemic stroke in the child, both before 30 days and before 365 days. The odds ratio point estimates were very similar to those for unconfirmed ischemic stroke, but were more strongly significant. Maternal diabetes was not associated with confirmed ischemic stroke in any of the models. The significant covariates for confirmed ischemic stroke were similar to those for any ischemic stroke.
Finally, we created a dichotomous variable (‘otherfactor’) to indicate that an infant was exposed to at least one of the following: birth trauma, birth asphyxia, thalassemia, sickle cell disease or trait, thrombophilia, congenital infection, neonatal infection, meningitis, encephalitis, or placental abruption. We then tested for multiplicative interactions between maternal hypertension and ‘otherfactor,’ and between maternal fever and ‘otherfactor’. Neither interaction term was significant when modeling the odds of ischemic stroke diagnosed before 30 days. When modeling stroke before 365 days, there was a significant interaction between maternal hypertension (but not fever) and the presence of ‘otherfactor’. When we stratified the models by the presence of another risk factor, we found that maternal hypertension was significantly associated with both any ischemic stroke (OR 2.46, 95% CI 1.58–3.84) and confirmed ischemic stroke (OR 2.72, 95% CI 1.66–4.46) before 365 days among children with at least one of the other risk factors. Maternal hypertension was not associated with any ischemic stroke or confirmed ischemic stroke before 365 days among children without any of the risk factors; the OR point estimates were approximately 1.0.
- Top of page
- What this paper adds
Maternal hypertension and intrapartum fever were both associated with neonatal ischemic stroke, whereas maternal diabetes was not. The findings related to hypertension are consistent with previous evidence that maternal pre-eclampsia is associated with increased risk of neonatal ischemic stroke. We are aware of only one other study23 identifying maternal fever as a significant risk factor for neonatal ischemic stroke. Intrapartum fever can be a marker for chorioamnionitis as well as for other maternal infections and is a risk factor for neonatal encephalopathy and death, cerebral palsy, and other adverse child outcomes.24,25 Intrapartum fever has previously been identified as a risk factor for neonatal death, neonatal encephalopathy, and several other adverse neonatal outcomes.25,26 Fever is associated with a significant systemic inflammatory response,27 and inflammatory cytokines are known to have prothrombotic effects that can precipitate ischemic events such as ischemic stroke.28,29
Although investigating non-maternal risk factors for neonatal ischemic stroke was not the primary goal of this study, it is interesting that we found multiple child characteristics to be associated with increased odds of neonatal/infant ischemic stroke. The significant covariates were generally related to increased propensity for thrombosis, were similar for neonatal ischemic stroke and for stroke up to 364 days after birth, and were consistent with risk factors previously identified for perinatal stroke.4 There appears to be a synergistic relationship between maternal hypertension and the presence of one or more other risk factors in predicting ischemic stroke diagnosed less than 365 days after birth. This synergy is consistent with the multifactorial model of perinatal ischemic stroke, which has been proposed by others.1,30 Taken as a whole, the variables examined in this study are good predictors of the risk of ischemic stroke in infants; among infants with ischemic stroke before 365 days, 65% were identified as having at least one of the risk factors (independent variable or covariate, not including preterm birth), compared with 22% of infants without ischemic stroke.
We also found that gestational age was substantially lower in infants with ischemic stroke. Most studies of perinatal/neonatal ischemic stroke stratify by preterm status and, therefore, cannot comment on the association between preterm birth and risk of ischemic stroke. However, Benders et al.31 report that, in a sample of 73 infants with perinatal arterial ischemic stroke, 42% were preterm (<37wks). We cannot necessarily conclude that preterm birth increases the risk of ischemic stroke. Instead, preterm birth may primarily be a marker of underlying problems in the pregnancy that are associated with increased risk of stroke. In-depth analysis of different causes of preterm birth and how they relate to risk of ischemic stroke in the infant would be needed to answer this question.
It has been postulated that perinatal stroke occurs because of vasculo-placental pathology such as infarctions, which lead to the formation of clots that can embolize to the fetal circulation, pass through the foramen ovale and into the cerebral vasculature.4,30 If this is the case, it stands to reason that maternal conditions that cause or are associated with placental infarction would be associated with increased risk of stroke in the infant. It is well established that placental infarction (and other lesions) are more common in pregnant females with hypertension.32,33 Dueck et al. 34 present a case study of an infant with perinatal stroke attributable to chorioamnionitis. They state, ‘In addition to the recognized inflammatory cascade of in utero infection, umbilical vein thrombosis with subsequent ‘paradoxical’ embolization may represent one mechanism responsible for this association.’ Additional research is needed to investigate likely mechanisms of the associations identified in our analyses.
The most important limitation of this study is its reliance on administrative data for ascertainment of both exposures and neonatal ischemic stroke. It is possible that some females identified as having fever and/or hypertension and some infants diagnosed with ischemic stroke were misclassified; however, given that the maternal diagnoses were made before the occurrence of stroke in the child, there is no reason to believe such misclassification would be differential for those exposures. Such a non-differential misclassification bias would tend to result in underestimation of the strength of the observed associations. It is important to note that, because perinatal ischemic stroke is estimated to occur in two to four infants per 10 000 live births, in our sample of almost 230 000 children, we would anticipate approximately 45 to 90 cases of neonatal ischemic stroke, which is similar to the range of confirmed cases identified in our cohort (37 cases diagnosed < 30d and 99 more between 30–365d).
A second limitation is that the females and children in this study were all enrolled in the South Carolina Medicaid program, which provides health insurance coverage for low-income families. Medicaid pays for approximately 50% of the births in South Carolina. It is possible that the results of this study may not be completely generalizable to middle- and upper-income families in South Carolina, or to females and children outside the southern USA.
A final limitation is that some strokes diagnosed between 30 days and 364 days may not be neonatal, but, rather, may have occurred later in infancy. Others30,35 have pointed out that up to 40 or 50% of neonatal strokes do not present until later in childhood, and that ischemic stroke is much more common in the perinatal period than at any other time in childhood.3 Despite this, it is likely that some of the strokes identified between 30 and 365 days actually did not occur during the neonatal period; we have no way to ascertain this with the information available. However, given that the 28-day cut-off for neonatal stroke is somewhat arbitrary, that strokes occurring later in infancy are also clinically significant, and that the risk factors we identified for stroke before 30 days were very similar to those for all strokes before 365 days, we believe it is appropriate to consider the two periods together as we have done.
The greatest strength of this study is its very large sample size, more than 200 000 births over 8 years, with linked maternal and child Medicaid, and birth certificate data. We had sufficient power to detect statistically significant risk factors for ischemic stroke. We were also able to control for several potential confounders and to perform sensitivity testing on variations in the case definition of neonatal ischemic stroke. Therefore, we believe it is appropriate to be highly confident in the findings of this study.
In summary, maternal hypertension and intrapartum fever appear to be risk factors for neonatal ischemic stroke. Additional research is needed to describe further the links between maternal hypertension and fever, and risk for neonatal stroke, and to investigate the mechanisms that may be involved. If these associations are confirmed, it may be feasible to develop targeted interventions to reduce the risk of ischemic stroke in infants born to females with these risk factors.