100

doi: 10.1111/j.1365-3016.2010.01157.x

Uncovering the complex relationship between pre-eclampsia, preterm birth and cerebral palsy
ppe_1157 100..110

Joshua R. Manna, Suzanne McDermotta, Margaret I. Griftha, James Hardinc and Anthony Greggb
Departments of aFamily and Preventive Medicine and bObstetrics and Gynecology, University of South Carolina School of Medicine, and
c

Department of Epidemiology and Biostatistics, University of South Carolina, Arnold School of Public Health, Columbia, SC, USA

Summary
Correspondence: Joshua R. Mann, Department of Family and Preventive Medicine, University of South Carolina School of Medicine, 3209 Colonial Drive, Columbia, SC 29203, USA. E-mail: joshua.mann@uscmed.sc.edu

Mann JR, McDermott S, Grifth MI, Hardin J, Gregg A. Uncovering the complex relationship between pre-eclampsia, preterm birth and cerebral palsy. Paediatric and Perinatal Epidemiology 2011; 25: 100110. Pre-eclampsia is a leading cause of preterm birth, which is strongly associated with cerebral palsy (CP). However, there is controversy about whether pre-eclampsia is associated with increased risk of CP. We evaluated the association between preeclampsia and CP in 122 476 motherchild pairs insured by the South Carolina Medicaid programme, with births between 1996 and 2002. Prenatal billing records were linked to the childrens Medicaid billing records after birth until December 2008. The odds of CP were modelled using logistic regression with generalised estimating equations. There were 337 children (0.28%) diagnosed with CP by at least two different health care providers, and 4226 (3.5%) women were diagnosed with pre-eclampsia at least twice during pregnancy. Children whose mothers had pre-eclampsia were almost twice as likely to have CP compared with children of mothers without pre-eclampsia [odds ratio (OR) = 1.94, 95% condence interval (CI) 1.25, 2.97]. The association was only signicant for pre-eclampsia diagnosed prior to 37 weeks gestation. Full term (gestational age 37 weeks) infants whose mothers were diagnosed with preeclampsia prior to 37 weeks had increased odds of CP compared with full term children whose mothers did not have pre-eclampsia (OR = 3.41, 95% CI 1.40, 8.31). Preterm infants whose mothers had pre-eclampsia were at signicantly increased risk of CP compared with full term infants whose mothers did not have pre-eclampsia (OR = 5.88, 95% CI 3.40, 10.17). The greatest risk for CP was in preterm infants whose mothers did not have pre-eclampsia (OR = 8.12, 95% CI 6.49, 10.17 compared with full term infants without exposure to pre-eclampsia). We conclude that pre-eclampsia with onset before 37 weeks gestation is a signicant risk factor for CP. Some of the association is probably attributable to high risk of preterm birth because of early preeclampsia, while a direct effect of pre-eclampsia on fetal brain development also seems likely. Keywords: pre-eclampsia, cerebral palsy, preterm delivery.

Introduction
Cerebral palsy (CP) is a permanent neurological condition characterised by abnormal motor function that originates early in life. Approximately 3 per 1000 children are affected by CP.1,2 Preterm delivery is strongly associated with risk of CP,3,4 and impaired

fetal growth is also a risk factor.2,5 While the epidemiological associations between preterm birth and impaired fetal growth are well established, it is important to keep in mind that these pregnancy outcomes do not occur in a vacuum but are themselves the results of abnormal physiological processes during pregnancy. For example, intrauterine infection appears to be an

Paediatric and Perinatal Epidemiology, 25, 100110. 2010 Blackwell Publishing Ltd.

Pre-eclampsia, cerebral palsy, preterm birth important underlying cause of both preterm birth and CP.3,4 Other risk factors for preterm birth and/or impaired fetal growth should likewise be considered potential causes of CP. Pre-eclampsia, dened as hypertension with proteinuria during pregnancy, occurs in approximately 5% of pregnancies.6 It is a known cause of intrauterine growth restriction.6,7 It is also a cause of preterm birth, although this effect is often iatrogenic (rather than physiological) as early delivery is the denitive treatment for severe pre-eclampsia.8 In fact, pre-eclampsia is a leading cause of iatrogenic preterm birth in the United States.9 Given the association between preeclampsia and both preterm birth and intrauterine growth restriction and the association of IUGR/ preterm delivery with CP, it is reasonable to hypothesise that pre-eclampsia may be associated with increased risk of CP. A number of studies have identied a signicant positive association between pre-eclampsia and CP.1012 However, those conducted in children born preterm have often found that pre-eclampsia was associated with lower risk of CP.1318 Based on these ndings, some have posited that pre-eclampsia actually has a protective effect.13,17 Greenwood et al. conducted a casecontrol study of pre-eclampsia and CP in 235 children with CP and 646 controls.19 The study demonstrated that while pre-eclampsia was inversely associated with CP in children born at <33 weeks gestation, there was no association in children born from 33 to 36 weeks gestation, and pre-eclampsia was associated with signicantly greater risk of CP in children born full term. The authors conclude that the apparent protective effect of pre-eclampsia in very preterm infants is probably not due to an actual decrease in risk but, rather, to the very high risk status of the control children born very preterm with no pre-eclampsia exposure. As the authors point out, Preterm birth is an abnormal event and all pathways that lead to preterm birth are pathological.19 The presence of reduced risk in preterm children born to women with preeclampsia may only mean that pre-eclampsia is not as big a risk factor for CP as some other causes of prematurity. One such condition that is known to cause both preterm birth and CP is intrauterine infection (often asymptomatic); this is believed to account for 25% 40% of all preterm births17 and up to 80% of extremely preterm births20,21 (where the risk of CP is dramatically elevated3,4). If preterm births in non-pre-eclamptic women are largely attributable to intrauterine infec-

101

tion, then it stands to reason that pre-eclampsia would appear to be protective in comparison. The ndings of Greenwood et al. provided substantial clarity about the relationship between preeclampsia and CP.19 However, additional questions remain. One important question is whether the timing of onset of pre-eclampsia inuences the association between pre-eclampsia and CP. This is an important question because only preterm pre-eclampsia can cause preterm birth, and because a number of authors have questioned whether early pre-eclampsia is in fact a clinical entity distinct from later pre-eclampsia.2224 A second related question is whether preterm infants born to women with early pre-eclampsia may be at increased risk of CP compared with low risk, full term infants. This information is vital if one wants to understand the overall contribution of pre-eclampsia to the risk of CP. The conceptual model for this study is shown in Figure 1. We hypothesise that early pre-eclampsia may increase the risk of CP by way of an increase in preterm delivery because of medically indicated preterm birth or by a direct effect on the fetal brain, or both. Late pre-eclampsia may increase the risk of CP only by way of a direct effect on brain development. Other pregnancy complications such as intrauterine infection can increase the risk of CP by way of preterm birth, by direct effects on the fetal brain, or both.

Methods
This project was granted exempt status by the University of South Carolina Institutional Review Board. We utilised a retrospective cohort study design that included de-identied South Carolina Medicaid billing records for pregnancies and deliveries that occurred between 1996 and 2002 inclusive. We also obtained linked les to birth certicates and Medicaid billing records for children. We limited our analysis to singleton pregnancies because multiple gestation is associated with adverse pregnancy outcomes including preterm birth and CP.25 We also excluded children diagnosed with a chromosome or genetic abnormality (Downs syndrome, Edwards syndrome, Patau syndrome, PraderWilli syndrome and Fragile X syndrome), a signicant anatomic anomaly of the brain, traumatic brain injury, cerebellar degeneration, central nervous system infections or other conditions (such as congenital infections) likely to result in abnormal brain development or brain injury that could cause or be

Paediatric and Perinatal Epidemiology, 25, 100110. 2010 Blackwell Publishing Ltd.

102

J. R. Mann et al.
Risk factors for pre-eclampsia

Figure 1. Conceptual model.

Early preeclampsia

Late preeclampsia

Medically indicated preterm delivery

Physiological effects of pre-eclampsia on fetal brain

Intrauterine infection; Other pregnancy complications

Spontaneous preterm labour

Preterm birth

Cerebral palsy

Physiological effects of infection and other pregnancy complications on fetal brain

confused with CP. In order to ensure adequate follow-up of infants, we limited our analyses to children who: 1. were diagnosed with CP, or 2. remained enrolled in South Carolina Medicaid until at least age 3, or 3. enrolled in the South Carolina Public School system. The outcome of CP was determined by identifying children diagnosed with CP in the Medicaid data, based on International Classication of Disease, Version 9 (ICD-9) code 343. We considered that some children could be misdiagnosed with CP and that children who truly had CP were likely to receive care from multiple providers (such as a primary care physician, a neurologist and/or a rehabilitation specialist). Therefore, we limited our analyses to conrmed cases of CP, dened as children who were diagnosed with CP by at least two different health care providers. Children diagnosed with CP by only one provider were excluded from analyses because of their uncertain case status. Pre-eclampsia/eclampsia (referred to throughout this article simply as pre-eclampsia) was identied based on ICD-9 codes 642.4642.7. To improve the validity of pre-eclampsia diagnoses, we required that an ICD-9 code indicative of pre-eclampsia be present in the Medicaid billing data at least twice during pregnancy. Women diagnosed with pre-eclampsia only once were excluded from analyses because of

their uncertain exposure status. We obtained an estimate of the gestational age at which pre-eclampsia was rst diagnosed, based on the date of rst diagnosis and gestational age at birth. For most analyses, timing of initial pre-eclampsia diagnosis was categorised as early (<37 completed weeks) or late (37 weeks or later). A small proportion of women with multiple diagnoses of pre-eclampsia were initially diagnosed prior to 20 weeks gestation; we are not certain whether these women truly had pre-eclampsia or not. We elected to retain them in the exposed group for the majority of the analyses, although we also tested the effect of restricting the analyses to women who were either diagnosed initially after 20 weeks or had no diagnosis of preeclampsia at all. We identied women who were diagnosed with a genito-urinary infection (in the Medicaid data) during the rst two trimesters, as we had previously identied early infection as a risk factor for CP26. Gestational age, maternal race and age, and childs sex were obtained from birth certicates. We excluded from all analyses infants with reported gestational age at delivery below 21 completed weeks, because (1) we were concerned that these very early deliveries could represent data entry errors, and (2) women who gave birth prior to 21 weeks had very little opportunity to be diagnosed with pre-eclampsia.

Paediatric and Perinatal Epidemiology, 25, 100110. 2010 Blackwell Publishing Ltd.

Pre-eclampsia, cerebral palsy, preterm birth

103

Statistical analyses
All analyses were performed in SAS 9.2. Descriptive statistics were tabulated, with signicance testing via the c2-test or Fishers exact test for categorical variables and Students t-test for continuous variables. The odds of CP were modelled using logistic regression. Because some women had more than one pregnancy during the study period, observations were not completely independent. Therefore, we modelled the outcome of a CP diagnosis using generalised estimating equations (the GENMOD procedure), with the binomial distribution and the logit link function where the denominator in the model was the number of pregnancies per mother, and the mother was the independent unit of analysis. We assumed an exchangeable correlation structure, in which the degree of correlation between pregnancies in the same mother remained constant over time. The independent variable for the initial analyses was pre-eclampsia diagnosed any time during pregnancy. However, for most of the analyses we looked beyond the presence or absence of a pre-eclampsia diagnosis during pregnancy to examine the impact of the timing of pre-eclampsia onset. We modelled specic months of pre-eclampsia diagnosis separately (that is, we estimated a separate model for each month of initial preeclampsia diagnosis, excluding observations where pre-eclampsia was initially diagnosed in a different month), but for the majority of the models, we exam-

ined early pre-eclampsia (diagnosed prior to 37 weeks gestation) vs. late pre-eclampsia (diagnosed at 37 weeks or later). The regression models were adjusted for maternal age and race (white, black or other), genito-urinary infection occurring in the rst two trimesters and childs sex. Because of the potential for preterm birth to lie in the causal pathway between early pre-eclampsia and CP, the primary models were estimated without controlling for gestational age or birthweight. We examined the interplay between pre-eclampsia and preterm birth (<37 weeks) by stratication. We conducted extensive sensitivity analyses to evaluate the robustness of the ndings. These secondary analyses focused on the effects of altering the inclusion/ exclusion criteria with respect to CP and pre-eclampsia diagnosis, and for high risk diagnoses in children.

Results
There were 145 048 births in the initial cohort. We excluded from analyses 3499 children who were not singleton births; 10 669 for whom the absence of CP could not be conrmed because they did not remain enrolled in Medicaid until at least age 3 and never enrolled in public schools; 5135 who were diagnosed with genetic/chromosomal abnormalities, traumatic brain injuries, central nervous system infections or

Table 1. Descriptive characteristics Children with cerebral palsy (n = 337) Variable Pre-eclampsia/ eclampsia Early PE (<37 weeks) Late PE (37 weeks) Non-Hispanic white non-Hispanic black Other <20 >35 2035 Category n 22 19 3 141 188 8 107 13 217 202 113 150 98 (%) (6.5) (5.6) (0.9) (41.8) (55.8) (2.4) (31.8) (3.9) (64.3) (59.9) (33.5) (44.5) (29.1) n 4204 2302 1902 55 880 64 747 1512 35 338 3480 83 321 62 179 31 483 11 381 2225 Children without cerebral palsy (n = 122 139) (%) (3.4) (1.9) (1.6) (45.8) (53.0) (1.2) (28.9) (2.9) (68.2) (50.9) (25.8) (9.3) (1.8) Pa 0.002 <0.0001 0.503 0.078

Race

Maternal age

0.241

Male sex GU infection rst two trimesters Preterm (<37 weeks) Very preterm (<33 weeks)
a

0.0009 0.001 <0.0001 <0.0001

Signicance testing using two-sided probabilities from Fishers exact test. GU, genito-urinary; PE, pre-eclampsia.

Paediatric and Perinatal Epidemiology, 25, 100110. 2010 Blackwell Publishing Ltd.

104

J. R. Mann et al.

Table 2. Multivariable modelling of preterm and very preterm birth

Timing of rst pre-eclampsia diagnosis No pre-eclampsia Any pre-eclampsia <37 weeks 20 weeks 2128 weeks 2936 weeks

OR [95% CI] for any preterm (<37 weeks) 1.0 Reference 5.4 [5.0, 5.8] 17.1 [15.6, 18.7] 2.5 [1.4, 4.7] 30.2 [22.8, 39.9] 18.2 [16.6, 20.0]

OR [95% CI] for very preterm (<33 weeks) 1.0 Reference 7.0 [6.3, 7.8] 14.2 [12.7, 16.0] 1.7 [0.4, 7.6] 127.8 [93.2, 159.2] 85.9 [71.9, 102.6]a

Models are adjusted for maternal age and race, genito-urinary infection in the rst two trimesters and childs sex. a Model for pre-eclampsia at 2936 weeks and very preterm birth is for pre-eclampsia diagnosed between 29 and 32 weeks only.

other conditions placing them at very high risk of neurological disability; 219 diagnosed with CP by only one health care provider; and 3033 whose mothers were diagnosed with pre-eclampsia/eclampsia only one time during pregnancy. Seventeen children were excluded because their reported gestational age was below 21 completed weeks. This left a sample of 122 476 for analysis. The prevalence of pre-eclampsia diagnosed at least twice was 3.5%. The sample is described in Table 1. Children with CP were signicantly more likely to have been born preterm, to be male and to have had a mother with genito-urinary infection diagnosed in the rst two trimesters. Race did not appear to be associated with CP status, and neither did maternal age. Of the 122 476 children 337 (0.28%) had conrmed CP. Pre-eclampsia was diagnosed in 22 (6.5%) mothers of children with conrmed CP, compared with 4204 (3.4%) of mothers of children without CP. Nineteen of the 22 cases of CP in children of women with pre-eclampsia were in infants born to women with early pre-eclampsia, and 14 of these 19 were in infants born prior to 37 weeks. There was no signicant difference in the proportion with late preeclampsia (37 weeks) for mothers of children with CP vs. those without. We modelled the odds of preterm (<37 weeks) and very preterm (<33 weeks) birth, to evaluate the extent to which pre-eclampsia was associated with these birth outcomes that are established risk factors for CP. Among the 2321 women with early pre-eclampsia, 1444 (62.2%) were born preterm (<37 weeks) and 448 (19.3%) were very preterm (<33 weeks). Adjusted odds ratios (ORs) are presented in Table 2. Pre-eclampsia/ eclampsia diagnosed at any time during pregnancy was associated with a vefold increase in the odds of

any preterm birth and a sevenfold increase in the odds of very preterm birth (<33 weeks). Pre-eclampsia diagnosed prior to 37 weeks gestation was even more strongly associated with preterm (OR = 17.1) or very preterm (OR = 14.2) birth. Women diagnosed with preeclampsia prior to 33 weeks gestation were >30 times as likely to deliver preterm, and >60 times as likely to deliver very preterm. These associations were highly signicant (P < 0.0001). As expected, gestational age was, in turn, strongly associated with the risk of CP. The odds of CP were signicantly increased among children born before 37 weeks [OR = 7.7, 95% condence interval (CI) 6.2, 9.7] and were even more substantially elevated in very preterm infants (OR = 22.2, 95% CI 17.1, 28.7) (data not shown in tabular form). Table 3 displays the risk of CP by timing of the initial diagnosis of pre-eclampsia, as well as adjusted ORs for CP by timing of pre-eclampsia diagnosis. There were 2321 women diagnosed with early pre-eclampsia and 1905 diagnosed at 37 weeks or later. The prevalence of CP was threefold as high in children of women with early pre-eclampsia (8.2 per 1000 children) as in children of women without pre-eclampsia (2.7 per 1000) (P < 0.0001). The overall incidence of CP in the sample was 2.8 per 1000. Based on these data, the attributable risk for CP among children whose mothers had preeclampsia was 5.5 per 1000, and the attributable risk percentage was 67.7%. The number of women with early pre-eclampsia required to produce one additional case of CP is estimated to be 181. The population attributable risk was 1 per 10 000, and early preeclampsia accounted for 3.8% of CP cases in the cohort (it should be kept in mind that children with high risk genetic, chromosomal or other conditions are not included in these calculations).

Paediatric and Perinatal Epidemiology, 25, 100110. 2010 Blackwell Publishing Ltd.

Pre-eclampsia, cerebral palsy, preterm birth

Table 3. Timing of pre-eclampsia and risk of CP Timing of rst pre-eclampsia diagnosis No pre-eclampsia Any pre-eclampsia <37 weeks 37 weeks <20 weeks 2024 weeks 2528 weeks 2932 weeks 3336 weeks

105

Total number 118 250 4226 2321 1905 57 63 196 565 1440

n with CP 315 22 19 3 0 5 3 3 8

n without CP 117 935 4204 2303 1902 57 58 193 562 1432

Proportion with CP 0.0027 0.0052 0.0082 0.0016 0.0000 0.0794 0.0153 0.0053 0.0056

Adjusted OR 1.0 Reference 1.9 [1.2, 3.0] 3.0 [1.9, 4.8] 1.0 [0.3, 3.0] 0.00 (NA)a 29.6 [11.7, 75.3] 5.8 [1.8, 18.4] 2.4 [0.8, 7.6] 2.9 [1.4, 5.8]

0.003 <0.0001 0.939 NAa <0.0001 0.003 0.129 0.004

Odds ratios are adjusted for maternal age and race, genito-urinary infection in the rst two trimesters and childs sex. The model for pre-eclampsia diagnosed at 37 weeks or later is limited to full term infants. a The condence interval and P-value are not meaningful, because of the lack of any cases of CP in the exposed group. CP, cerebral palsy; NA, not applicable.

The risk of CP was generally greater the earlier preeclampsia was diagnosed. There were 57 women diagnosed with pre-eclampsia prior to 20 completed weeks (40 of these 57 were prior to 17 weeks); these cases of pre-eclampsia are dubious as pre-eclampsia is dened as occurring after at least 20 weeks of completed gestation.27 None of the women diagnosed with preeclampsia prior to 20 completed weeks of gestation had a child with CP. However, the rate of CP was 7.9% when pre-eclampsia was diagnosed between 20 and 24 weeks gestation and 1.5% when pre-eclampsia was diagnosed between 25 and 28 weeks. There was no signicant difference in the rate of CP for children of women with pre-eclampsia diagnosed at 37 weeks or later, compared with non-pre-eclamptic women who delivered full term. The unadjusted OR for CP in children of women with pre-eclampsia (either early or late) was 2.0 (95% CI 1.3, 3.0). For early pre-eclampsia, the unadjusted OR was 3.1 (95% CI 2.0, 5.0), while for late pre-eclampsia it was 0.9 (95% CI 0.3, 2.9). The association between preeclampsia and CP remained when adjusted for maternal age and race, genito-urinary infection in the rst two trimesters and childs sex. Overall, children of women with pre-eclampsia were approximately twice as likely to have CP (OR = 1.9, 95% CI 1.3, 3.0). The association was entirely accounted for by preeclampsia diagnosed prior to 37 weeks (OR = 3.0, 95% CI 1.9, 4.8), while pre-eclampsia diagnosed at term was not associated with CP (OR = 1.0, 95% CI 0.3, 3.0). We modelled the impact of early pre-eclampsia using more precise categories for timing of rst diagnosis. In

each of these models, we included only cases of preeclampsia diagnosed in that specic time window and excluded infants born before the beginning of the time window. For example, in evaluating the impact of preeclampsia diagnosed between 29 and 32 weeks, infants born before 29 weeks were excluded. Odds of CP were increased 30-fold in children of women diagnosed with pre-eclampsia between 20 and 24 weeks compared with children whose mothers were not diagnosed with preeclampsia (OR = 29.6, 95% CI 11.7, 75.3). We stratied by preterm birth status and modelled the odds of CP, early pre-eclampsia and late preeclampsia being the independent variables. This modelling, shown in Table 4, revealed that early preeclampsia was signicantly associated with increased risk of CP in full term infants (OR = 3.5, 95% CI 1.4, 8.4) but not in preterm infants (OR = 0.7, 95% CI 0.4, 1.3). Late pre-eclampsia was not signicantly associated with odds of CP (OR = 1.0, 95% CI 0.3, 3.0). As described in the Introduction, preterm infants exposed to pre-eclampsia may have a lower risk of CP than children who are preterm because of other disease processes (such as intrauterine infection), yet still be at increased risk compared with infants whose mothers did not have pre-eclampsia, given that pre-eclampsia is a cause of preterm birth. To better understand the role of pre-eclampsia as a risk factor for CP in the context of preterm and full term birth, we created a ve-level variable to express combinations of pre-eclampsia and preterm/full term status, and then modelled the odds of CP. The ve levels of the variable were: (1) early pre-eclampsia, preterm birth, (2) early pre-eclampsia,

Paediatric and Perinatal Epidemiology, 25, 100110. 2010 Blackwell Publishing Ltd.

106

J. R. Mann et al.

Table 4. Adjusted odds ratios for CP, stratied by preterm status

First pre-eclampsia diagnosis Any pre-eclampsia <37 weeks 37 weeks

OR for CP in preterm infants NA 0.7 [0.4, 1.3] NA

P NA 0.256 NA

OR for CP in term infants 1.7 [0.9, 3.5] 3.5 [1.4, 8.4] 1.0 [0.3, 3.0]

P 0.124 0.006 0.939

Models are adjusted for maternal age and race, genito-urinary infection in the rst two trimesters and childs sex. The model for pre-eclampsia diagnosed at 37 weeks or later is limited to full term infants. CP, cerebral palsy; NA, not applicable.

full term birth, (3) no pre-eclampsia, preterm birth, (4) late pre-eclampsia, full term birth, and (5) no pre-eclampsia, full term birth. The fth level (no preeclampsia, full term) served as the reference group. The model was adjusted for maternal age and race, genitourinary infection in the rst two trimesters, and childs sex. The results are shown in Table 5. The highest risk of CP was in preterm infants born to women without pre-eclampsia (OR = 8.1, 95% CI 6.5, 10.2). However, infants born to women with early pre-eclampsia also had signicantly increased risk whether they were preterm (OR = 5.9, 95% CI 3.4, 10.2) or full term (OR = 3.4, 95% CI 1.4, 8.3). Late pre-eclampsia was not associated with risk of CP. We performed extensive sensitivity analyses to test the robustness of the ndings. First, we tested the effect of excluding the 57 women diagnosed with preeclampsia prior to 20 weeks of completed gestation, and repeated the model shown in Table 4. Full term infants whose mothers had early pre-eclampsia were again at signicantly increased risk compared with

Table 5. Pre-eclampsia-preterm cerebral palsy Pre-eclampsia-preterm status No pre-eclampsia, full term birth Early pre-eclampsia, preterm Early pre-eclampsia, full term No pre-eclampsia, preterm Late pre-eclampsia, full term

status

and

odds

of

OR 1.0 5.9 3.4 8.1 1.0

[95% CI] Reference [3.4, 10.2] [1.4, 8.3] [6.5, 10.2] [0.3, 3.0]

<0.0001 0.007 <0.0001 0.936

Model is adjusted for maternal age and race, genito-urinary infection in the rst two trimesters and childs sex.

full term controls (OR = 3.6, 95% CI 1.5, 8.8), as were preterm infants whose mothers had early preeclampsia (OR = 5.9, 95% CI 3.4, 10.2). Late preeclampsia in full term infants was again not signicant (OR = 0.9, 95% CI 0.3, 2.9). Preterm infants whose mothers did not have pre-eclampsia had a higher OR than preterm infants with pre-eclamptic mothers, but the CIs overlapped (OR = 8.1, 95% CI 6.5, 10.2). Next, we removed the requirement that women be diagnosed with pre-eclampsia on at least two different occasions to be included in the exposed group. This increased the number of women with pre-eclampsia to 7259 (5.8%). We repeated the model from Table 4, which again demonstrated that compared with full term controls, full term infants whose mothers had early pre-eclampsia were at signicantly increased risk (OR = 2.9, 95% CI 1.4, 5.8), as were preterm infants whose mothers had early pre-eclampsia (OR = 6.2, 95% CI 3.9, 9.9). Full term infants whose mothers had late pre-eclampsia were not at signicantly increased risk (OR = 1.5, 95% CI 0.8, 2.9). Preterm infants whose mothers did not have pre-eclampsia had the highest risk (OR = 8.1, 95% CI 6.5, 10.1). Next, we estimated the model without excluding children who had documented genetic, chromosomal or other conditions that placed them at high risk of CP, independent of pre-eclampsia or gestational age. Even when including these children in the model, preterm (OR = 3.8, 95% CI 2.3, 6.4) and full term infants (OR = 2.5, 95% CI 1.1, 5.8) whose mothers had early pre-eclampsia had signicantly increased odds of CP compared with full term infants not exposed to preeclampsia. Preterm infants whose mothers did not have pre-eclampsia were also at signicantly increased risk (OR = 5.6, 95% CI 4.6, 6.9). Late pre-eclampsia was not associated with increased odds of CP (OR = 0.6, 95% CI 0.2, 1.7). We tested the effect of including all children diagnosed with CP by at least one provider in the outcome

Paediatric and Perinatal Epidemiology, 25, 100110. 2010 Blackwell Publishing Ltd.

Pre-eclampsia, cerebral palsy, preterm birth group. This increased the number of cases to 539. Both preterm (OR = 4.5, 95% CI 2.8, 7.1) and full term (OR = 2.7, 95% CI 1.3, 5.7) children whose mothers had early pre-eclampsia were more likely to be diagnosed with CP than full term children with non-preeclamptic mothers. Late pre-eclampsia was again not signicant (OR = 1.1, 95% CI 0.5, 2.4). Once again, preterm infants whose mothers did not have preeclampsia had the highest risk (OR = 6.4, 95% CI 5.3, 7.7). We were concerned that earlier diagnosis of preeclampsia might be a proxy for more severe disease, so we estimated the model, limiting the sample to women who were diagnosed only with mild or unspecied pre-eclampsia (ICD-9 code 642.4). This resulted in the removal of 250 women with severe pre-eclampsia, eclampsia or pre-eclampsia/eclampsia super-imposed on pre-existing hypertension. Preterm infants with pre-eclamptic mothers were at signicantly increased risk (OR = 6.4, 95% CI 3.7, 11.0), as were preterm infants with non-pre-eclamptic mothers (OR = 8.1, 95% CI 6.5, 10.2) and full term infants exposed to early pre-eclampsia (OR = 2.8, 95% CI 1.1, 7.7). Late preeclampsia was not signicantly associated with odds of CP (OR = 0.7, 95% CI 0.2, 2.6). Finally, we re-estimated the model using what we considered to be the most conservative approach. We did not make any exclusions based on high risk diagnoses in the children, and we included all women with any diagnosis of pre-eclampsia as exposed and all children with any diagnosis of CP as cases. The only exclusions were non-singleton infants, those with reported gestational age <21 weeks, and lack of sufcient follow-up in the Medicaid or public school les. Having a diagnosed medical exclusion was included as a covariate in the model. Overall, early pre-eclampsia was signicantly associated with CP (OR = 2.0, 95% CI 1.5, 2.7). As expected, children with a medical exclusion diagnosis were far more likely to have CP (OR = 12.3, 95% CI 10.3, 14.6). When the ve-level variable for combined pre-eclampsia-preterm status was modelled as the independent variable, preterm infants with pre-eclamptic mothers (OR = 3.6, 95% CI 2.5, 5.1) and preterm infants with non-pre-eclamptic mothers (OR = 4.6, 95% CI 3.9, 5.4) were at signicantly increased risk of CP. Full term infants whose mothers had early pre-eclampsia were not at signicantly increased risk, although the point estimate was >1.0 (OR = 1.7, 95% CI 1.0, 2.9). Late pre-eclampsia was not signicant (OR = 1.1, 95% CI 0.7, 1.8).

107

Discussion
This analysis produced several key ndings. First, preeclampsia was a signicant predictor of increased risk of CP when timing of pre-eclampsia diagnosis was not considered. Second, the earlier pre-eclampsia was diagnosed, the stronger the association with CP appeared to be. Pre-eclampsia diagnosed before 37 weeks was strongly associated with increased odds of CP, with diagnosis at 2024 weeks or 2528 weeks associated with particularly high risk. Third, pre-eclampsia diagnosed before 37 weeks was a signicant risk factor for CP in full term infants in almost all the models. Finally, while preterm infants whose mothers had preeclampsia were not at increased risk of CP compared with other preterm infants, the risk of CP in preterm infants with pre-eclamptic mothers was far greater than the risk in full term controls. Pre-eclampsia diagnosed at 37 weeks or later was not signicantly associated with odds of CP, although it must be noted that the OR estimate lacked precision because of small cell size; there were only three cases of CP in full term infants whose mothers were diagnosed with pre-eclampsia at 37 weeks or later. In addition, our decision to limit the primary analyses to women with at least two separate diagnoses of pre-eclampsia resulted in the exclusion of many women who were diagnosed with pre-eclampsia at the end of their pregnancy and did not have an opportunity to be diagnosed again. Including the women with a single diagnosis of pre-eclampsia increased the OR point estimate for late pre-eclampsia to 1.50, which still was not statistically signicant. The ndings were robust in extensive sensitivity testing. The sensitivity test that had the greatest impact was the inclusion of children with genetic/ chromosomal abnormalities or other high risk conditions. Point estimates for early pre-eclampsia also declined when relaxing the criteria for being exposed to pre-eclampsia and for being a case of CP. Only in the most conservative approach (inclusion of any diagnosis of pre-eclampsia and CP, without excluding children diagnosed with high risk conditions), was the increased odds of CP for early pre-eclampsia among full term infants not statistically signicant; even in this case the 95% CI [0.96, 2.91] narrowly extended below 1.0, providing some evidence of increased risk. It stands to reason that relaxing the exposure and outcome criteria would result in a diminution of the observed associations, as non-differential misclassica-

Paediatric and Perinatal Epidemiology, 25, 100110. 2010 Blackwell Publishing Ltd.

108

J. R. Mann et al. of each preterm birth risk factor from the risk of immaturity per se; therefore, questions about the mechanism of brain injury in preterm children of pre-eclamptic mothers may ultimately prove unanswerable in human studies. It is possible that some unobserved characteristic that is associated with both early pre-eclampsia and CP may account for the association between early pre-eclampsia and CP in full term infants. For example, there could be genetic or physiological characteristics of the mother or fetus that increase the risk of both pre-eclampsia and CP. While the cause of preeclampsia is unknown, it is generally accepted that abnormal placental development plays a signicant role,28 and research is accumulating to support the important role of placental function in the aetiology of CP29. The nature of our study does not permit us to rule out the possibility of confounding or common causation; nor does it provide information on the likely mechanism(s) by which early pre-eclampsia may cause CP. Basic science research is needed to explain how pre-eclampsia diagnosed early in pregnancy may inuence fetal brain development. In recent years a number of researchers have identied what appeared to be a protective effect of preeclampsia related to CP when comparisons are made between two groups of preterm infants: one with mothers who had pre-eclampsia and one with mothers who did not have pre-eclampsia.13,1518 As pointed out by Greenwood et al.,19 in actuality this protective effect was probably the result of the decision to stratify by preterm birth or to limit the study to preterm infants. Hernndez-Daz et al.30 and Whitcomb et al.31 describe the phenomenon of collider-stratication bias that can occur when stratifying or controlling for birthweight (which is, of course, closely related to immaturity) in perinatal research. The problem, as described by Whitcomb and colleagues, is that when two variables share a common effect, stratication on the variable representing that effect induces a statistical relationship between otherwise independent factors. In the case of research on pre-eclampsia, preterm birth and CP, the most important confounding factor is probably intrauterine infection, which is a major risk factor for both preterm birth and CP. By stratifying by, adjusting for or restricting a sample to children with preterm birth, a researcher creates a comparison group that probably has a much higher rate of intrauterine infection (as well as other risk factors for preterm birth that

tion of exposures and outcomes is likely to produce a bias towards the null. Similarly, it is likely that cerebral palsy diagnosed in the context of genetic or chromosomal abnormalities, brain injuries, CNS infections and other high risk conditions is fundamentally distinct from cases occurring in the absence of any of these overwhelming risk factors. If so, including these cases would also tend to produce a bias towards the null. We conducted an additional analysis, limiting the sample to children who had been diagnosed with one of the medical exclusions. In this model, neither early nor late pre-eclampsia was associated with odds of CP, and the association between preterm birth without preeclampsia and CP was much weaker than in our primary analysis (OR = 2.8, vs. 8.1 in the primary analysis). The much weaker effect of preterm birth in children with high risk conditions supports the assertion that CP diagnosed in children with these conditions may not be equivalent to CP diagnosed in the remainder of the cohort. Therefore, we believe our initial approach of excluding these children was appropriate. Based on the increased risk of CP in full term infants whose mothers had early pre-eclampsia, it appears that the contribution of early pre-eclampsia to CP risk includes a direct effect on the fetal brain that is not accounted for by increased risk of preterm birth. At the same time, the high rate of preterm birth in the context of early pre-eclampsia appears to be very important. Sixty-two per cent of women with early pre-eclampsia delivered preterm infants, and the risk of CP in these children was higher than in full term infants, although it was non-signicantly lower than in other preterm infants. We believe it is appropriate to infer that the majority of preterm deliveries in women with preeclampsia are attributable to pre-eclampsia, and therefore pre-eclampsia is the underlying cause of most cases of CP in these infants. For this reason, controlling for preterm status by stratication or inclusion of gestational age as a covariate is not appropriate if the goal is to gain an estimate of the total contribution of pre-eclampsia to the risk of CP. Without access to clinical information about details surrounding preterm births in both pre-eclamptic and non-pre-eclamptic women, such as information about asymptomatic intrauterine infection or other pregnancy complications, it is impossible to know whether pre-eclampsia affects the risk of CP in preterm infants beyond the effects of prematurity itself. Even if full information on the cause of every preterm birth were available, it would be very difcult to isolate the impact

Paediatric and Perinatal Epidemiology, 25, 100110. 2010 Blackwell Publishing Ltd.

Pre-eclampsia, cerebral palsy, preterm birth may also be associated with CP) than is present in the general population of pregnant women. This results in an elevated risk of CP in the controls and thereby reduces the observed effect of pre-eclampsia, as most pre-eclamptic women who deliver preterm probably do not have these other, unmeasured risk factors for preterm birth and CP. This is not to say that there is no justication for comparing the risk of CP for preterm infants with pre-eclamptic mothers with those with non-pre-eclamptic mothers; there may be valid reasons to do this, such as to provide accurate prognostic information for clinicians to provide to the families of preterm infants. However, if the goal is to test whether pre-eclampsia is a potential cause of CP and to quantify the overall contribution of pre-eclampsia to CP risk, we believe the most appropriate approach is to avoid stratifying by, adjusting for or conditioning on preterm birth. One important difference in our study compared with most previous ones is that we compared the risk of CP in children whose mothers were diagnosed with pre-eclampsia with the entire group of children whose mothers did not have pre-eclampsia and then to full term infants whose mothers did not have preeclampsia, thereby capturing the total contribution of pre-eclampsia to CP risk compared with what the anticipated risk would have been if their mothers had not had pre-eclampsia. Furthermore, we examined the role of timing of onset of pre-eclampsia, which is important as (1) only relatively early onset preeclampsia has the potential to cause CP by way of preterm birth, (2) previous research indicates that susceptibility to periventricular leukomalacia (one of the most frequent pathological ndings related to CP) appears to be greatest between 27 and 30 weeks of gestation,32 and (3) others have proposed that early and late pre-eclampsia may be fundamentally different entities.2224 Therefore, it stands to reason that preeclampsia developing early in pregnancy could have a different (and probably greater) impact on risk of CP from pre-eclampsia than onset late in pregnancy. Our study is limited by its utilisation of administrative data (Medicaid billing records) for identifying women with pre-eclampsia and children with CP. We believe that most cases of pre-eclampsia were properly identied as blood pressure and urinanalyses are routinely monitored during prenatal care. Of course women with minimal or no prenatal care could have had their pre-eclampsia missed if it was mild, as mild pre-eclampsia is typically asymptomatic. Our restric-

109

tion of the cohort to women who were either diagnosed with pre-eclampsia on two separate occasions or were never diagnosed with pre-eclampsia reduces the likelihood of misdiagnosed cases of pre-eclampsia, although the associations generally changed only slightly when adding the women with one-time diagnoses of pre-eclampsia back into the models. There may have been some misdiagnoses of CP in the children, but again our decision to limit the analyses to children who were diagnosed with CP by at least two providers or were not diagnosed at all minimises this risk. Adding back children with CP diagnosed by only one provider reduced the strength of the associations (which would be expected in the event of nondifferential misclassication bias), but the changes were fairly small and probably not clinically signicant. A nal limitation is that Medicaid is an insurance programme for low income women and children, thus the ndings are not necessarily applicable to all income groups. Additional epidemiological research in more economically diverse populations is needed to determine whether the ndings are generalisable. In conclusion, early pre-eclampsia is associated with increased risk of CP. Part of the increased risk appears to be attributable to the very high rate of preterm birth in women with early pre-eclampsia, and 14 of 19 cases of CP in children of women with early pre-eclampsia were delivered preterm. Unfortunately, we are not able to determine whether there is a direct effect of early pre-eclampsia on CP risk in preterm infants, as there is no way of knowing the degree to which preterm birth in a completely normal pregnancy would increase CP risk. However, it appears likely that there is also a direct association between early pre-eclampsia and CP in full term infants. Additional research is needed to examine the biological mechanisms whereby pre-eclampsia may alter fetal brain development.

Acknowledgements
We would like to acknowledge Pete Bailey and Heather Kirby from the South Carolina Ofce of Research and Statistics for providing the data used in these analyses, and Haikun Bao from the University of South Carolina, Arnold School of Public Health for his assistance with data analysis. The study was funded by the Health Resources and Services Administration, #R40MC11277-01-00.

Paediatric and Perinatal Epidemiology, 25, 100110. 2010 Blackwell Publishing Ltd.

110

J. R. Mann et al.
16 Tm OS, Kothasdia JM, Roberts DD, Dillard RG. Perinatal events and the risk of intraparenchymal echodensity in very low birthweight neonates. Paediatric and Perinatal Epidemiology 1998; 12:408421. 17 Spinillo A, Capuzzo E, Cavallini A, Stronati M, De Santolo A, Fazzi E. Preeclampsia, preterm delivery and infant cerebral palsy. European Journal of Obstetrics, Gynecology, and Reproductive Biology 1998; 77:151155. 18 OShea TM, Klinepeter KL, Dillard RG. Prenatal events and the risk of cerebral palsy in very low birth weight infants. American Journal of Epidemiology 1998; 147:362369. 19 Greenwood C, Yudkin P, Sellers S, Impey L, Doyle P. Why is there a modifying effect of gestational age on risk factors for cerebral palsy? Archives of Disease in Childhood. Fetal and Neonatal Edition 2005; 90:F141F146. 20 Findings from the 199298 Patient Outcomes Research Team on Low Birthweight. Clinical Highlight. AHRQ Publication No. 00-P010, January 2000. Agency for Healthcare Research and Quality, Rockville, MD. http://www.ahrq.gov/clinic/ lobrhigh.htm [last accessed July 2010]. 21 Goldenberg RL, Culhane JF, Iams JD, Romero R. Epidemiology and causes of preterm birth. Lancet 2008; 371:7584. 22 Vatten LJ, Skjaerven R. Is pre-eclampsia more than one disease? BJOG 2004; 111:298302. 23 Sibai B, Dekker G, Kupferminc M. Pre-eclampsia. Lancet 2005; 365:785799. 24 Masuyama H, Segawa T, Sumida Y, Masumoto A, Inoue S, Akahori Y, et al. Different proles of circulating angiogenic factors and adipocytokines between early- and late-onset pre-eclampsia. BJOG 2010; 117:314320. 25 Moore AM, OBrien K. Follow-up issues with multiples. Paediatrics & Child Health 2006; 11:283286. 26 Mann JR, McDermott S, Bao H, Bersabe A. Maternal genitourinary infection and risk of cerebral palsy. Developmental Medicine and Child Neurology 2009;51:282288. 27 Leeman L, Fontaine P. Hypertensive disorders of pregnancy. American Family Physician 2008; 78:93100. 28 Kanasak K, Kalluri R. The biology of preeclampsia. Kidney International 2009; 76:831837. 29 Redline RW. Disorders of placental circulation and the fetal brain. Clinics in Perinatology 2009; 36:549559. 30 Hernndez-Daz S, Schisterman EF, Hernn MA. The birth weight paradox uncovered? American Journal of Epidemiology 2006; 164:11151120. 31 Whitcomb BW, Schisterman EF, Perkins NJ, Platt RW. Quantication of collider-stratication bias and the birthweight paradox. Paediatric and Perinatal Epidemiology 2009; 23:394402. 32 Zupan V, Gonzalez P, Lacaze-Masmonteil T, Boithias C, dAllest AM, Dehan M, et al. Periventricular leukomalacia: risk factors revisited. Developmental Medicine and Child Neurology 1996; 38:10611067.

References
1 Yeargin-Allsopp M, Van Naarden Braun K, Doernberg NS, Benedict RE, Kirby RS, Durkin MS. Prevalence of cerebral palsy in 8-year-old children in three areas of the United States in 2002: a multisite collaboration. Pediatrics 2008; 121:547554. 2 Bhasin TK, Brocksen S, Avchen RN, Van Naarden Braun K. Prevalence of four developmental disabilities among children aged 8 years Metropolitan Atlanta Developmental Disabilities Surveilance Program, 1996 and 2000. MMWR. Surveillance Summaries 2006; 55:19. 3 Clark SM, Ghulmiyyah LM, Hankins GDV. Antenatal antecedents and the impact of obstetric care in the etiology of cerebral palsy. Clinical Obstetrics and Gynecology 2008; 51:775786. 4 Nelson KB. Causative factors in cerebral palsy. Clinical Obstetrics and Gynecology 2008; 51:749762. 5 Jacobsson B, Ahlin K, Francis A, Hagberg G, Hagberg H, Bardosi J. Cerebral palsy and restricted growth status at birth: population-based case-control study. British Journal of Obstetrics and Gynaecology 2008; 115:12501255. 6 Duley L. The global impact of pre-eclampsia and eclampsia. Seminars in Perinatology 2009; 33:130137. 7 Chammas MF, Nguyen TM, Li MA, Nuwayhid BS, Castro LC. Expectant management of severe preterm preeclampsia: is intrauterine growth restriction an indication for immediate delivery? American Journal of Obstetrics and Gynecology 2000; 183:853858. 8 Leeman L, Fontaine P. Hypertensive disorders of pregnancy. American Family Physician 2008; 78:93100. 9 Ananth CV, Vintzileos AM. Medially indicated preterm birth: recognizing the importance of the problem. Clinics in Perinatology 2008; 35:5367. 10 Ozturk A, Demirci F, Yavuz T, Yildiz S, Degirmenci Y, Dosoglu M, et al. Antenatal and delivery risk factors and prevalence of cerebral palsy in Duzce (Turkey). Brain & Development 2007; 29:3942. 11 Thorngren-Jerneck K, Herbst A. Perinatal factors associated with cerebral palsy in children born in Sweden. Obstetrics and Gynecology 2006; 108:14991505. 12 Anwar S, Chowdhury J, Khatun M, Mollah AH, Begum HA, Rahman Z, et al. Clinical prole and predisposing factors of cerebral palsy. Mymensingh Medical Journal 2006; 15:142 145. 13 Xiong X, Saunders LD, Wang FL, Dadge ST, Buekens P. Preeclampsia and cerebral palsy in low birth weight and preterm infants: implications for the current ischemic model of preeclampsia. Hypertension in Pregnancy 2001; 20:113. 14 Gray PH, Jones P, OCallaghan MJ. Maternal antecedents for cerebral palsy in extremely preterm babies: a case-control study. Developmental Medicine and Child Neurology 2001; 43:580585. 15 Wilson-Costello D. Risk factors for neurologic impairment among very low birth weight infants. Seminars in Pediatric Neurology 2001; 8:120126.

Paediatric and Perinatal Epidemiology, 25, 100110. 2010 Blackwell Publishing Ltd.