Preterm delivery is a major risk factor for perinatal death and neurodevelopmental disability, including cerebral palsy, in surviving newborns.1 – 3 It has been hypothesized that cerebral palsy and other forms of neurodevelopmental disability are the result of neuronal injury through inflammatory, hypoxic, excitatory, or oxidative mechanisms. Seventy percent of cases are now believed to be attributable to prenatal or perinatal factors, with birth asphyxia only playing a minor role in those born preterm.2,4 Antenatal exposure to magnesium sulfate before anticipated early preterm delivery has been shown to reduce the risk of cerebral palsy and motor dysfunction among survivors.5 – 7 The final clinical outcome of at-risk children probably is affected by postnatal events and interactions with the postnatal environment.1
Genetic susceptibility may modify an individual's risk for adverse outcomes both before and after delivery.1,8 There is evidence that certain genetic polymorphisms in inflammation pathways (eg, cytokine polymorphisms) or thrombosis pathways (eg, inherited thrombophilias) may be associated with a greater risk of cerebral palsy and neurodevelopmental delay.8,9 There are limited data regarding the role of genetic polymorphisms in other pathways, particularly those involved with other mechanisms of neuronal injury.8 Therefore, our objective was to evaluate the association between neurodevelopmental diability at age 2 or death and polymorphisms in candidate genes involved in neuronal homeostasis and protection or oxidative stress, and to assess whether these genetic polymorphisms modulate the neuroprotective effects of magnesium sulfate.
PATIENTS AND METHODS
This was a secondary analysis and a nested case-control study of the Eunice Kennedy Shriver National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network Randomized Clinical Trial of the Beneficial Effects of Antenatal Magnesium Sulfate, in which women with singleton or twin gestations between 24 and 31 6/7 weeks of gestation at high risk for imminent preterm birth were randomized in a double-blind fashion to receive either magnesium sulfate infusion or identical-appearing placebo.5 Group assignment in the original trial was made according to a computer-generated random sequence using the urn design, with stratification according to clinical center and, in twin pregnancies, weeks of gestation (less than 28 weeks or 28 weeks or more).10 Further details of the study, which was conducted at 20 institutions between December 1997 and May 2004, are described elsewhere.5 The primary outcome of the trial was a composite of infant death or moderate to severe cerebral palsy at the age of 2 years. Additionally, as secondary outcomes, the trial assessed other neurodevelopmental outcomes (psychomotor and mental delay) at age 2.
For the current analysis, we included children who died by 1 year of corrected age or had cerebral palsy, mental delay, or psychomotor delay diagnosed at or beyond 2 years of corrected age. DNA and 2-year data on neurologic outcomes were required for inclusion. Mental and psychomotor delay were defined by scores less than 70 at age 2 years using the Bayley Scales of Infant Development II mental and psychomotor developmental indices administered by a centrally certified psychologist or psychometrist. Cerebral palsy was determined by a centrally certified pediatrician or pediatric neurologist according to prespecified criteria of gross motor delay, abnormality in muscle tone or movement, or reflexes.5 The control group consisted of children who survived until 2 years of age with normal neurodevelopment (Bayley mental developmental index and psychomotor developmental index score 85 or more) and without cerebral palsy, periventricular leukomalacia, or intraventricular hemorrhage grade III or IV. We excluded cases of fetal demise and children with major congenital malformations or genetic syndromes. Randomly selected control group children were matched to case group children for ethnicity or race and infant sex to minimize the chance of misidentifying a race-based or sex-based genotype as one associated with the outcome of interest. Control group children were matched in a ratio of 4:1 to children in the case group with cerebral palsy, 2:1 to children in the case group experiencing cerebral palsy or death, and 1:1 to children in the case group with mental and psychomotor delays. This study was deemed exempt from Institutional Review Board review because the data and samples were deidentified before the analysis was performed.
The DNA extraction, amplification, and genotyping were performed at Taueret Laboratories (Salt Lake City, UT). Researchers and laboratory personnel were blinded to the case status and control status of the biologic samples. The DNA was extracted from umbilical cord serum that had been collected at the time of delivery using the PureGene DNA Purification System, per the manufacturer's protocols and as previously described, and then amplified for better yield.9 Serum samples were the only biologic material available for analysis. Because of DNA degradation and subsequent variation in size of polymerase chain reaction products, DNA extracted from serum is often of poorer quality compared with DNA extracted from whole blood. This can result in a larger number of missing genotype values when evaluating several single-nucleotide polymorphisms (SNPs), which can result in discrepant performance among SNP assays. Therefore, we excluded children whose samples either failed to genotype or had more than 30% of the genotype results missing (n=47; 23 in the case group and 24 in the control group), and then we excluded their corresponding case and control group participants to maintain the matching design. We also randomly selected one twin from each pair of included twins (n=6) to avoid including related individuals in the analysis. Triplets and higher-order multiple gestations were excluded from the primary trial. We selected 21 SNPs in 17 genes associated with oxidative stress and neuronal homeostasis and neuroprotection based on the available literature and hypothesized causal pathways.11 – 26 The SNPs that were included in our custom multiplex assay are shown in Table 1. The SNP analysis was performed using the Taqman SNP allele discrimination assay for sample genotyping. Predesigned TaqMan assays were used for genotyping, and genotypes were determined using Applied Biosystems automated Taqman genotyping software SDS 2.1.
Demographic characteristics between cases with and without DNA samples available were compared using χ2, Fisher exact, or Wilcoxon rank-sum tests for maternal characteristics and generalized estimating equations for neonatal characteristics to account for the correlation between siblings.27 Demographic characteristics between case group and control group participants were compared using χ2, Fisher exact, or Wilcoxon rank-sum test. Outcomes evaluated were cerebral palsy, the combined outcome of cerebral palsy or death, mental delay, and psychomotor delay. The combination of cerebral palsy or death is necessary because cerebral palsy and death are competing outcomes. The former cannot be assessed when the latter occurs before a reliable diagnosis of cerebral palsy can be made. Each SNP was tested for Hardy Weinberg equilibrium in the case group participants (because control group participants were matched to case group participants in this study). Logistic regression modeling was used to analyze the association between each SNP and neonatal outcomes accounting for demographic and clinical variables significantly different between case and control group participants or known a priori to be associated with neurodevelopmental outcomes. This included maternal education, which is associated with the results of the Bayley Scales of Infant Development (the source of our measure of psychomotor and mental delay), gestational age at delivery, which is associated with the various neurodevelopmental outcomes, and magnesium sulfate, which has been shown to be associated with a decrease in cerebral palsy. For each outcome–SNP combination, we first checked the interaction effect between the SNP and treatment assignment. If the interaction term was not significant, then we removed it from the model and focused on the SNP effect only. Genotypes were included as covariates in the regression model that assumed an additive genetic pattern. The additive genetic model assumes that each polymorphism confers additional risk and that having two copies of the minor allele has twice the effect of having one copy, and has been preferred for its robustness among the other genetic models. All the models were based on the logistic regression conditional on maternal race (African American, white, or Hispanic) and neonatal sex (male or female). Because this was an exploratory study, no power calculation was conducted, and no adjustments were made for multiple comparisons. A two-sided P<.05 was considered to indicate statistical significance. Statistical analyses were performed using SAS statistical software.
The primary data collection was approved by the Institutional Review Boards of the Biostatistical Coordinating Center and the clinical sites at which patients were recruited. All women provided written informed consent.
The original trial (Beneficial Effects of Antenatal Magnesium Sulfate) included 2,444 infants, of whom 2,260 infants had long-term neurodevelopmental outcomes available. Of those, 668 infants had cerebral palsy diagnosed, mental or psychomotor delay diagnosed, or they had died, and they satisfied our definition of “cases.” However, of those 668 infant case group individuals, only 231 had DNA available to be analyzed. These then were matched to 231 infant “controls” by maternal ethnicity or race and infant sex. After excluding infant samples that failed to genotype or had more than 30% of the genotype missing (n=47; 23 case group participants and 24 control group participants), and after randomly selecting one twin from each pair (n=6; 5 case group participants and 1 control group participants), the cohort remaining for analysis included 409 infants (203 case group participants and 206 control group participants) (Fig. 1).
Compared with the 425 infant case group participants from the parent cohort who satisfied the inclusion criteria but had no available DNA, the 203 infant case group participants with DNA samples available for testing were more likely to be born singleton and at a more advanced gestational ages (Table 2).
For this analysis, there were 43 infants who died, 24 children with cerebral palsy, 109 with mental delay, and 91 with psychomotor delay. There were eight children with cerebral palsy and mental delay, 14 with cerebral palsy and psychomotor delay, and 48 with mental and psychomotor delay. The maternal and neonatal characteristics of all case group participants and control group participants are summarized in Table 3. Compared with control group participants, case group participants were born at an earlier gestational age. Those with mental delay were born to mothers with less education than the matched control group participants, and those with cerebral palsy were less likely to be from singleton gestations. There were no differences in ethnicity, study drug assignment, chorioamnionitis, or infant sex between case group participants and control group participants (Table 3).
Out of 21 SNPs chosen, the assay for one SNP (Neuregulin-1, NRG1, rs35753505) failed manufacturing, and another SNP (N-methyl-D-aspartate receptor 3B subunit, GRIN3B C/T, rs480739) failed to genotype in more than 30% of the samples, and these were excluded from further analysis. In the remainder, genotype determination was successful overall in 94% of samples (range 88%–99% for individual SNPs). Four SNPs were in Hardy Weinberg disequilibrium (P<.01): vasoactive intestinal polypeptide (VIP) rs17083008; glutathione peroxidase 1 (GPX1) rs1800668; reelin (RELN) rs362691; and vasoactive intestinal polypeptide receptor 2 (VPAC2) rs885861.
The (VIP rs17083008 and N-methyl-D-aspartate receptor 3A subunit [GRIN3A] rs3739722) SNPs were associated with cerebral palsy. The genotype frequency of these SNPs in case group participants and control group participants are summarized in Table 4. The VIP (A) allele and GRIN3A (T) allele were associated with increased risk of cerebral palsy (adjusted odds ratio [OR] 2.67, 95% confidence interval [CI] 1.09–6.55, P=.03, and adjusted OR 4.67, 95% CI 1.36–16.01; P=.01, respectively). We tried to add an additional interaction term but observed no significant interaction with magnesium sulfate treatment (P=.86 and 0.89, respectively). The advanced glycosylation end product–specific receptor (AGER) rs3134945 SNP was differentially associated with mental delay, depending on exposure to magnesium sulfate. For this SNP there was an interaction with magnesium sulfate treatment (P=.02); therefore, we analyzed the association in the placebo and magnesium sulfate–treated groups separately. Despite this significant interaction, however, there was no significant association between the minor allele (A) and mental delay either in the placebo group (adjusted OR 0.51, 95% CI 0.24–1.07, P=.08) or in the magnesium sulfate group (adjusted OR 1.83, 95% CI 0.89–3.73, P=.10) (Fig. 2).
In this study, we found that polymorphisms in the VIP and GRIN3A genes were associated with an increased risk of cerebral palsy. The AGER rs3134945 SNP was differentially associated with mental delay, depending on exposure to magnesium sulfate.
Vasoactive intestinal polypeptide is an endogenous neuromodulator and neuroprotective peptide in the central nervous system.28 Its neuroprotective role against ischemic and glutamate-induced excitatory brain injury is mediated by vasoactive intestinal peptide receptors (VPAC1 and VPAC2 subtypes).14,15 Vasoactive intestinal peptide stimulates astrocytes to produce neuroprotective agents, particularly the potent activity-dependent neurotrophic factor and activity-dependent neuroprotective protein, and inhibits the production of inflammatory cytokines by activated microglia.28 Functional genetic polymorphisms in both VIP and VPAC2 genes have been described.14 – 16 In particular, the VIP rs17083008, located in the promoter or regulatory region, is predicted to have a potential functional role because it alters a transcription factor–binding site, may be associated with lower vasoactive intestinal polypeptide levels, and is thought to be associated with bipolar and other neurobehavioral disorders.16
The N-methyl-D-aspartate receptors are tetrameric glutamate channels expressed on astrocytes and oligodendrocytes and are essential in excitatory synaptic transmission in the central nervous system.29,30 It has been postulated that N-methyl-D-aspartate receptors play an essential role in the pathogenesis of cerebral palsy.31,32 Briefly, in the setting of hypoxic injury, cells shift to anaerobic metabolism and increase lactate and glutamate production. The latter then binds and coactivates N-methyl-D-aspartate receptors, leading to influx of Ca++ intracellularly, lipid peroxidation, free radical production, and cell death or injury. Magnesium is a noncompetitive inhibitor of these N-methyl-D-aspartate receptors blocking the influx of Ca++ into the cells and thus protects against cell damage or death.31,32 There are three major subunits of the N-methyl-D-aspartate tetramer (NR1, NR2, and NR3).29 The N-methyl-D-aspartate receptor 3A subunit is one of the two types of the NR3 subunit and has been shown to be highly expressed in oligodendrocytes, where glutamate toxicity leads to damage in the myelin sheath. Its expression in the dendirtic spines implies a role in synaptic stability and neuronal activity.29,30 The N-methyl-D-aspartate receptor 3A subunit is abundantly expressed on the myelin sheath and is only weakly blocked by extracellular Mg2+, making it more vulnerable to glutamate toxicity damage.29 The currently described SNP is an exonic polymorphism at position 3723 in the GRIN3A gene that encodes the N-methyl-D-aspartate receptor 3A subunit and has been found to be associated with Alzheimer disease.22
Last, animal and human studies suggest that inflammation and oxidative stress in the developing fetal brain are associated with brain injury and subsequent development of cerebral palsy.31 – 33 The host's defense to an inflammatory insult involves a group of intracellular proteins called damage-associated molecular pattern molecules that, when released into the extracellular milieu after an insult, activate the receptor for advanced glycosylation end products.33 This will then lead to activation of multiple cellular signaling cascades mediated by nuclear factor κB, and to amplification of the inflammatory and oxidative stress signaling, potentiating and accelerating cellular damage and dysfunction.33 We previously have shown in a nested case-control study from the same cohort that a naturally occurring soluble truncated variant of the receptor for advanced glycosylation end products was decreased in the cord blood of children who ultimately died after preterm birth,34 indicating a possible role of receptor for advanced glycosylation end products activation in neurodevelopmental disability or death. In this study, we have found that the AGER SNP (rs3134945), which is located in the 3′ untranslated region of the advanced AGER gene and its function is still unknown, was differentially associated with mental delay, depending on exposure to magnesium sulfate. However, despite this significant interaction, there was no significant association between the minor allele (A) and mental delay in either group.
The major strength of this study is that it was nested in a multicenter trial in which the outcomes were clearly defined and the data and specimen were carefully collected in a standard fashion. Umbilical cord serum samples were the only biologic material available for analysis. Because of the poorer quality of DNA extracted from serum, we eliminated individual serum samples and specific SNP assays that genotyped poorly compared with other samples and SNPs. We then excluded their corresponding case or control participant to maintain the matching status. We used an elimination cut-off of more than 30% missing results for serum samples and SNP assays. Given the relatively small number of candidate SNPs, a more rigorous cut-off would have resulted in exclusion of a large number of high-quality serum samples with as few as one missing genotype. After excluding serum samples that entirely failed to genotype (n=26) and samples with more than 30% of genotypes missing (n=21), the average individual sample genotype call rate over the 19 markers was 98%. Using the same exclusions, the 19 SNPs had an average genotype call rate of 97%. This elimination method for both individual serum samples and specific SNP assays resulted in a robust genetic analysis without bias toward inclusion of erroneous genotypes. Our study was limited by the sample size that did not permit us to evaluate gene–environment interactions on subsequent survival and neurodevelopmental outcomes.
Table 2 compares infant case group participants with or without DNA available, and it was intended to see whether the remaining sample is representative of the entire study cohort. We speculated that the difference in gestational age and twin gestations seen between infants who had DNA samples available, compared with those who did not, may be related to difficulty in obtaining sufficient cord blood from the smallest neonates and in multiple gestations. These differences are not a significant source of systematic selection bias for the case-control secondary analysis reported, although they may limit ability to extrapolate to the primary study population. We also performed an analysis with a covariate for twin compared with singleton gestation, and all results have remained similar. In addition, data were not available to assess the influence of other variables that may be associated with adverse neurodevelopmental outcomes such as histologic chorioamionitis, funisitis, and acidosis at birth. Because this was an exploratory analysis, no corrections were made for multiple comparisons and all associations were reported despite the possibility that some associations may be statistically significant but not clinically meaningful, and some statistically significant associations may have been attributable to chance (type I error). Our findings need to be confirmed in other cohorts. The possibility of false-negatives cannot be excluded as well. Given the limited number of SNPs analyzed, the genetic variation within the selected genes was not fully captured.
In conclusion, survival and risk of adverse neurodevelopmental outcomes are determined by complex interactions between genes and the intrauterine and postnatal environment. VIP and GRIN3A SNPs may be associated with cerebral palsy at age 2 in children who were born preterm. This work supports the potential role of genetic predisposition to death and neurodevelopmental disability in preterm infants and suggests that genetic predisposition may alter responses to antenatal magnesium sulfate treatment and risk of death and adverse neurodevelopmental outcomes. Knowledge of individual fetal genetic risk ultimately may lead to new preventive and therapeutic strategies that optimize neurodevelopment after preterm birth.
1. American College of Obstetricians and Gynecologists, American Academy of Pediatrics. Neonatal encephalopathy and cerebral palsy: defining the pathogenesis and pathophysiology. Washington, DC: ACOG; 2003.
2. Paneth N, Hong T, Korzeniewski S. The descriptive epidemiology of cerebral palsy. Clin Perinatol 2006;33:251–67.
3. Saigal S, Doyle LW. An overview of mortality and sequelae of preterm birth from infancy to adulthood. Lancet 2008;371:261–9.
4. Hagberg B, Hagberg G, Beckung E, Uvebrant P. Changing panorama of cerebral palsy in Sweden: VIII. Prevalence and origin in the birth year period 1991–94. Acta Paediatr 2001;90:271–7.
5. Rouse DJ, Hirtz DG, Thom E, Varner MW, Spong CY, Mercer BM, et al.. A randomized, controlled trial of magnesium sulfate for the prevention of cerebral palsy. N Engl J Med 2008;359:895–905.
6. Doyle LW, Crowther CA, Middleton P, Marret S, Rouse D. Magnesium sulphate for women at risk of preterm birth for neuroprotection of the fetus. The Cochrane Database of Systematic Reviews 2009, Issue 1. Art. No.: CD004661. DOI: 10.1002/14651858/CD004661.pub3.
7. Costantine MM, Weiner SJ, Eunice Kennedy Shriver
National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network. Effects of antenatal exposure to magnesium sulfate on neuroproetction and mortality in preterm infants: a meta-analysis. Obstet Gynecol 2009;114:354–64.
8. O'Callaghan ME, MacLennan AH, Haan EA, Dekker G, South Australian Cerebral Palsy Research Group. The genomic basis of cerebral palsy: a HuGE systematic literature review. Hum Genet. 2009;126:149–72.
9. Clark EA, Mele L, Wapner RJ, Spong CY, Sorokin Y, Peacman A, et al.. Association of fetal inflammation and coagulation pathway gene polymorphisms with neurodevelopmental delay at age 2 years. Am J Obstet Gynecol 2010;203:83.e1–10
10. Wei LJ, Lachin JM. Properties of the urn randomization in clinical trials. Controlled Clin Trials 1988;9:345–64.
11. Kleffner I, Bungeroth M, Schiffbauer H, Schäbitz WR, Ringelstein EB, Kuhlenbäumer G. The role of aquaporin-4 polymorphism in the development of brain edema after middle cerebral artery occlusion. Stroke 2008;39:1333–5.
12. Sorani MD, Manley GT, Giacomini KM. Genetic variation in human aquaporins and effects on phenotypes of water homeostasis. Hum Mutat 2008;29:1108–17.
13. Dai Q, Shrubsole MJ, Ness RM, Schlundt D, Cai Q, Smalley WE, et al.. The relation of magnesium and calcium intakes and a genetic polymorphism in the magnesium transporter to colorectal neoplasia risk. Am J Clin Nutr 2007;86:743–51.
14. Delgado M, Robledo G, Rueda B, Varela N, O'Valle F, Hernandez-Cortes P, Caro M, Orozco G, et al.. Genetic association of vasoactive intestinal peptide receptor with rheumatoid arthritis altered expression and signal in immune cells. Arthritis Rheumatism 2008;58:1010–9.
15. Sun W, Hong J, Zang YC, Liu X, Zhang JZ. Altered expression of vasoactive intestinal peptide receptors in T lymphocytes and aberrant Th1 immunity in multiple sclerosis. Int Immunol 2006;18:1691–700.
16. Soria V, Martinez-Amoros E, Escaramis G, Valero J, Perez-Egea R, Garcia C, et al.. Differential association of circadian genes with mood disorders: CRY1 and NPAS2 are associated with unipolar major depression and CLOCK and VIP with bipolar disorder. Neuropscychopharmacology 2010;35:1279–89.
17. Cheeran B, Talelli P, Mori F, Koch G, Suppa A, Edwards M, et al.. A common polymorphism in the brain-derived neurotrophic factor gene (BDNF) modulates human cortical plasticity and the response to rTMS. J Physiol 2008;586:5717–25.
18. Serajee FJ, Zhong H, Mahbubul Huq AH. Association of Reelin gene polymorphisms with autism. Genomics 2006;87:75–83.
19. Krug T, Manso H, Gouveia L, Sobral J, Xavier JM, Albergaria I, et al.. Kalirin: a novel genetic risk factor for ischemic stroke. Hum Genet 2010;127:513–23.
20. Hoffmann I, Bueter W, Zscheppang K, Brinkhaus MJ, Liese A, Riemke S, et al.. Neuregulin-1, the fetal endothelium, and brain damage in preterm newborns. Brain Behav Immun 2010;24:784–91.
21. Li D, He L. Association study between the NMDA receptor 2B subunit gene (GRIN2B) and schizophrenia: A HuGE review and meta-analysis. Genet Med 2007;9(1):4–8.
22. Liu HP, Lin WY, Liu SH, Wang WF, Tsai CH, Wu BT, et al.. Genetic variation in N-methyl-D-aspartate receptor subunit NR3A but not NR3B influences susceptibility to Alzheimer's disease. Dement Geriatr Cogn Disord 2009;28:521–7.
23. Gibson CS, MacLennan AH, Dekker GA, Goldwater PN, Sullivan TR, Munroe DJ, et al.. Candidate genes and cerebral palsy: a population-based study. Pediatrics 2008;122:1079–85.
24. Rajaraman P, Hutchinson A, Rothman N, Black PM, Fine HA, Loeffler JS, et al.. Oxidative response gene polymorphisms and risk of adult brain tumors Neuro Oncol 2008;10:709–15.
25. Gaens KHJ, Van Der Kallen CJH, Van Greenvenbroek MMJ, et al.. Receptor for advanced glycation end product polymorphisms and type 2 diabetes. The CODAM study. Ann NY Acad Sci 2008;1126:162–5.
26. Arbour NC, Lorenz E, Schutte BC, Zabner J, Kline JN, Jones M, et al.. TLR4 mutations are associated with endotoxin hyporesponsiveness in humans. Nat Genet 2000;25:187–91.
27. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73:13–22.
28. Dejda A, Sokolowska P, Nowak JZ. Neuroprotective potential of three neuropeptides PACAP, VIP and PHI. Pharmacol Rep 2005;57:307–20.
29. Mayer ML. Glutamate receptors at atomic resolution. Nature 2006;440:456–62.
30. Karadottir R, Cavelier P, Bergersen LH, Attwell D. NMDA receptors are expressed in oligodendrocytes and activated in ischaemia. Nature 2005;438:1162–6.
31. Alvarez-Diaz A, Hilario E, Goni de Cerio FG, Valls-i-Soler A, Alvarez-Diaz FJ. Hypoxic-ischemic injury in the immature brain–key vascular and cellular players. Neonatology 2007;92:227–35.
32. Costantine MM, Drever N. Antenatal exposure to magnesium sulfate and neuroprotection in preterm infants. Obstet Gynecol Clin North Am 2011;38:351–66, xi.
33. Buhimschi CS, Baumbusch MA, Dulay AT, Oliver EA, Lee S, Zhao G, et al.. Characterization of RAGE, HMGB1, and S100beta in inflammation-induced preterm birth and fetal tissue injury. Am J Pathol 2009;175:958–75.
© 2012 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.
34. Costantine MM, Weiner SJ, Rouse DJ, Hirtz DG, Varner MW, Spong CY, et al.. Umbilical cord blood biomarkers of neurologic injury and the risk of cerebral palsy or infant death. Int J Dev Neurosci. 2011;29:917–22.