search text   Advanced Search

This Article Abstract Full Text (PDF) P3Rs: Submit a response Alert me when this article is cited Alert me when P3Rs are posted Alert me if a correction is posted Citation Map Services E-mail this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Add to My File Cabinet Download to citation manager Citing Articles Citing Articles via CrossRef Citing Articles via ISI Web of Science (7) Citing Articles via Google Scholar Google Scholar Articles by Ambalavanan, N. Search for Related Content PubMed PubMed Citation Articles by Ambalavanan, N. Related Collections Premature & Newborn

Predicting Outcomes of Neonates Diagnosed With Hypoxemic-Ischemic Encephalopathy

^a Department of Pediatrics, University of Alabama at Birmingham, Birmingham, Alabama
^b Department of Pediatrics, Wayne State University, Detroit, Michigan
^c RTI International, Research Triangle Park, North Carolina
^d National Institute of Child Health and Human Development Neonatal Research Network, Bethesda, Maryland
^e Department of Pediatrics, University of Texas Health Science Center at Houston, Houston, Texas
^f Department of Pediatrics, Wake Forest University School of Medicine, Winston-Salem, North Carolina
^g Department of Pediatrics, Women and Infants Hospital, Providence, Rhode Island
^h Department of Pediatrics, Yale University School of Medicine, New Haven, Connecticut
ⁱ Department of Pediatrics, University of Cincinnati, Cincinnati, Ohio
^j Department of Pediatrics, Case Western Reserve University, Cleveland, Ohio
^k Department of Pediatrics, Duke University, Durham, North Carolina

	ABSTRACT

TOP ABSTRACT METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
OBJECTIVE. The goals were to identify predictor variables and to develop scoring systems and classification trees to predict death/disability or death in infants with hypoxic-ischemic encephalopathy.

METHODS. Secondary analysis of data from the multicenter, randomized, controlled, National Institute of Child Health and Human Development Neonatal Research Network trial of hypothermia in hypoxic-ischemic encephalopathy was performed. Data for 205 neonates diagnosed as having hypoxic-ischemic encephalopathy were studied. Logistic regression analysis was performed by using clinical and laboratory variables available within 6 hours of birth, with death or moderate/severe disability at 18 to 22 months or death as the outcomes. By using the identified variables and odds ratios, scoring systems to predict death/disability or death were developed, weighting each predictor in proportion to its odds ratio. In addition, classification and regression tree analysis was performed, with recursive partitioning and automatic selection of optimal cutoff points for variables. Correct classification rates for the scoring systems, classification and regression tree models, and early neurologic examination were compared.

RESULTS. Correct classification rates were 78% for death/disability and 71% for death with the scoring systems, 80% and 77%, respectively, with the classification and regression tree models, and 67% and 73% with severe encephalopathy in early neurologic examination. Correct classification rates were similar in the hypothermia and control groups.

CONCLUSIONS. Among neonates diagnosed as having hypoxic-ischemic encephalopathy, the classification and regression tree model, but not the scoring system, was superior to early neurologic examination in predicting death/disability. The 3 models were comparable in predicting death. Only a few components of the early neurologic examination were associated with poor outcomes. These scoring systems and classification trees, if validated, may help in assessments of prognosis and may prove useful for risk-stratification of infants with hypoxic-ischemic encephalopathy for clinical trials.

Key Words: logistic models • decision trees • asphyxia neonatorum • hypoxia-ischemia • brain • predictive value of tests

Abbreviations: CART—classification and regression tree • GMFCS—gross motor function classification system • HIE—hypoxic-ischemic encephalopathy • ROC—receiver operating characteristic • NICHD—National Institute of Child Health and Human Development • CI—confidence interval

Hypoxic-ischemic encephalopathy (HIE) is a major contributor to neonatal death and morbidity. An estimated 23% of the 4 million neonatal deaths¹ and 8% of all deaths at <5 years of age throughout the world each year are associated with signs of asphyxia at birth.² Even at referral centers in developed countries, death or moderate to severe disability occurs for 53% to 61% of infants diagnosed as having moderate to severe HIE.^3,4 Statements about prognoses in HIE are usually based on neurologic examinations. Sarnat and Sarnat⁵ described 3 clinical stages of encephalopathy, and they and others have reported that infants with mild (stage 1) encephalopathy rarely have persistent neurologic impairment, as seen for a majority of infants with severe (stage 3) encephalopathy. Infants with moderate encephalopathy have an intermediate risk of persistent neurologic impairment.^5–7 Children with moderate/severe neonatal encephalopathy are at risk for reduced school performance, whereas those with mild encephalopathy have school performance scores similar to those of their peers.⁶ Despite the fact it has been shown to be highly predictive of outcomes,^5–7 one limitation of the staging described by Sarnat and Sarnat⁵ is that the stage is usually assigned well after birth; no studies have evaluated the predictive value of early neurologic evaluations. It is extremely important to validate the use of neurologic examinations in the early hours after resuscitation for recruitment for therapeutic trials and to examine whether additional clinical data can strengthen predictions.

A recent randomized, controlled trial of hypothermia in infants with HIE, conducted by the National Institute of Child Health and Human Development (NICHD) Neonatal Research Network, enrolled 208 term infants and demonstrated reductions in rates of death and moderate/severe disability, with an absolute risk reduction of 18% (number needed to treat = 6) and a relative risk reduction of 29%.³ This trial provided a unique opportunity to develop a scoring system, because this was a multicenter trial with a relatively large sample size from tertiary care NICUs in recent years. In addition, all of the enrolled infants had neurologic evaluations performed by certified physician examiners before enrollment, detailed histories, and follow-up evaluations performed by certified examiners at 18 to 22 months of age. Because hypothermia reduces rates of death and morbidity attributable to HIE and hypothermia needs to be initiated soon after birth to be effective,^8–10 it is of clinical importance to ascertain prognosis soon after birth and, if possible, to identify a subset of neonates for whom hypothermia may be more beneficial.

Our objectives were to identify predictor variables and to develop scoring systems and prognostic algorithms for HIE in term infants and to compare their predictive values with that of severe encephalopathy, as diagnosed through the early neurologic examinations used in the NICHD Neonatal Research Network hypothermia trial. A secondary goal was to determine the precooling attributes that alone or in combination could predict infants for whom hypothermia might be of greater benefit.

	METHODS

TOP ABSTRACT METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
Subjects
This study was a secondary analysis of data from the NICHD Hypothermia Trial.³ Infants were screened if they were of gestational age of 36 weeks and were admitted to the NICU at <6 hours of age, with either poor respiratory effort at birth and a need for resuscitation or a diagnosis of encephalopathy. Infants were evaluated according to physiologic criteria and then through a neurologic examination. Eligibility criteria included pH of 7.0 or base deficit of 16 mmol/L in cord blood or during the first 1 hour after birth. If, during this interval, the pH was between 7.01 and 7.15, the base deficit was between 10 and 15.9 mmol/L, or an arterial blood gas value was not available, then additional criteria were required (an acute perinatal event and either a 10-minute Apgar score of 5 or assisted ventilation for 10 minutes from birth). Once these criteria were met, all infants underwent a standardized neurologic examination (Table 1), performed by a certified physician examiner. Encephalopathy was defined as 1 sign in at least 3 of the 6 categories. The number of moderate or severe signs determined the degree of encephalopathy; if signs were distributed equally, then the designation was based on the level of consciousness.³ Moderate or severe encephalopathy or seizures qualified infants for the trial.³

Model Development
Complementary approaches of logistic regression analysis and classification and regression tree (CART) analysis were used for development of the scoring systems and prognostic algorithms, respectively. Clinical and laboratory variables available at the time of initial evaluation (<6 hours of age), including components of the early neurologic evaluation, were used for development of the models (Table 2). Separate logistic regression analysis with stepwise selection of variables (at P < .1) identified variables in the entire cohort associated with death/moderate/severe disability and death (SAS 9.0; SAS Institute, Cary, NC). The variables identified by stepwise regression were used to develop a logistic regression model for predicting the outcome of interest. The odds ratios from this logistic regression model were converted into point totals of >1 by dividing each odds ratio by the smallest odds ratio for any given level of a variable. Next, an overall score was assigned to each infant by summing the points they received for their various levels of each variable in the model. An infant with a missing value for any source variable was given a missing value for the overall score, because omission of an item would result in a lower score. Models with imputed values for the missing values were also tested but were not used because they did not demonstrate better accuracy. Continuous variables (eg, base deficit) were divided into 3 categories by developing a receiver operating characteristic (ROC) curve for the variable with respect to the outcome of interest and identifying the cutoff points for 80% sensitivity and 80% specificity. These scoring systems were applied to the entire cohort, the hypothermia cohort, and the control cohort separately. On the basis of the mutually exclusive ranges of scores, we identified (1) the score below which all infants had good outcomes (absence of either death/moderate/severe disability or death, depending on the model) even without hypothermia, (2) the score range in which infants might have benefited from hypothermia (all infants in the hypothermia group had good outcomes but some infants in the control group had poor outcomes), (3) the score range in which the benefit is less clear (poor outcomes were noted for some control infants and infants in the hypothermia group), and (4) the score above which all infants died or were disabled despite hypothermia.

	RESULTS

TOP ABSTRACT METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
Study Group
Of the 205 infants with primary outcome data, 172 infants (excluding 33 infants with missing data for the selected predictor variables) could be analyzed for the outcome of death/disability and 171 (excluding 34 with missing data) could be analyzed for the outcome of death. Missing data were mainly variables from the first postnatal blood gas analyses, which were not recorded despite other variables from those blood gas analyses being recorded. Infants with missing data, after adjustment for center, were comparable to those with available data with respect to nonmissing variables. Of the 172 infants analyzed for death or disability, 88 (51%) either died or were diagnosed as having disability; of the 171 infants analyzed for the outcome of death, 54 (32%) died.

Scoring Systems
The predictor variables, odds ratios, and scores assigned to each variable for death/disability scoring are shown in Table 3, and those for death scoring are shown in Table 4. For death/disability, the most influential variables, in order of highest odds ratios were decerebrate posture, base deficit of >22 mmol/L in first postnatal blood gas analysis, 5-minute Apgar score of <4, absence of spontaneous activity, and absence of chronic hypertension/pregnancy-induced hypertension. For death, the most influential variables were base deficit of >22 mmol/L in first postnatal blood gas analysis, decerebrate or distal flexion posture, absence of suck reflex, and absence of maternal antepartum hemorrhage.

CART Models
The CART model for death/disability is shown in Fig 1 and that for death in Fig 2. The infant's clinical data determine the path down the decision tree that is to be taken, and the incidence of the outcome in the terminal node or box is the probability of the outcome for the infant. The correct classification rates were improved significantly for death/disability but not for death, compared with early neurologic examination, as determined by the 95% CIs (Tables 5 and 6). The performances of the CART models and the logistic regression-based scoring systems were comparable. The correct classification rates were comparable in the hypothermia and control groups (Tables 5 and 6). Predictive variables in order of importance (higher on the decision tree) for death/disability included cord pH of 6.70 (91% death/morbidity) vs >6.70 (46%) (Fig 1). Among infants with cord pH of >6.70, those without spontaneous activity had worse outcomes (76% death/disability), compared with those with such activity (32% death/disability). Infants with some (normal/decreased) activity had worse outcomes if the base deficit in the first arterial blood gas analysis was >18.5 mmol/L (55% vs 18% for 18.5 mmol/L) (Fig 1). Predictors for death were base deficit in the first postnatal arterial blood gas analysis of >24.5 mmol/L (72% death) vs 24.5 mmol/L (21%). For infants with base deficits of 24.5 mmol/L, a cord pH of <6.74 predicted worse outcomes (48% death vs 15% for pH of >6.74) (Fig 2).

View larger version (27K): [in this window] [in a new window]   FIGURE 1 CART model for death/disability. The dichotomous outcome of death or disability at the 18- to 22-month follow-up assessment is predicted by this decision tree. In each node (rectangle), the category (Cat.) 0 or 1 refers to the absence or presence of death/disability, respectively, and the percentages and n values refer to the infants in each of the categories. The increment in positive predictive value with the variable under consideration is shown as the improvement (eg, 0.09 = 9% increase in positive predictive value). ABG indicates arterial blood gas.

View larger version (23K): [in this window] [in a new window]   FIGURE 2 CART model for death. The outcome of death by the 18- to 22-month follow-up assessment is predicted by this decision tree. In each node (rectangle), the category (Cat.) 0 or 1 refers to the absence or presence of death, respectively, and the percentages and n values refer to the infants in each of the categories. The increment in positive predictive value with the variable under consideration is shown as the improvement (eg, 0.078 = 7.8% increase in positive predictive value). ABG indicates arterial blood gas.

	DISCUSSION

TOP ABSTRACT METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

Strengths of this study include the relatively large sample, recruited from multiple sites, and prospective data collection by trained observers, both at birth and at follow-up evaluations. In addition, rather than empirical or expert opinion-derived scoring systems, we used variables associated statistically with the outcomes of interest and cutoff points for continuous variables that optimized discrimination between those with and without these outcomes.

A limitation is that, despite the relatively large sample size for this population, it was not feasible to use a split-half cross-validation approach in which we could develop the model in half of the data set and test it in the other half. Therefore, these scoring systems and decision trees need to be validated with other data sets. Also, as with any clinical trial, only the variables that were considered to be important for the purposes of the trial were collected, and other potentially important variables were not collected in the first 6 hours (eg, biochemical variables such as serum or cerebrospinal fluid lactate levels, electroencephalographic or cerebral blood flow recordings, or other clinical, laboratory, imaging, or special investigation data). It must also be emphasized that none of the models had correct classification rates close to 100%. Therefore, these models are suitable for risk stratification or assessment of prognosis but should not be a basis for decisions regarding withdrawal of support. Infants evaluated in this study experienced a severe insult, were admitted to tertiary care NICUs within 6 hours after birth, and fulfilled strict eligibility criteria for entry into the hypothermia trial. It is likely that infants who do not have similar characteristics at admission have different outcomes, and the predictive value of the decision trees and scoring systems may be substantially different when applied to such infants. In any prognostic system, certain variables (eg, pH and PCO₂) are relatively objective, whereas others, such as those noted in clinical examinations (eg, seizures), may be somewhat subjective, despite certification of examiners, leading to significant intercenter and inter-rater variations and variations in predictive ability across centers or clinicians. It is probable that the accuracy of prediction could be improved by adding variables that are determined later in the clinical course (eg, results of cranial imaging or clinical variables assessed closer to discharge). However, one of the aims of our study was to determine whether we could identify infants who would benefit from hypothermia, which needs to be initiated soon after birth to be effective¹⁰; therefore, we used only early variables. Our models are suitable for rapid assessment of sick neonates, because they rely mostly on objective measures such as pH and base deficit, together with a few well-defined clinical findings that are available earlier and may promote earlier identification and initiation of treatment. A detailed neurologic or radiologic evaluation is often difficult to perform for sick neonates, for whom issues of stabilization (eg, airway and intravenous access) take precedence.

A comparison of the scoring systems based on logistic regression and the CART models reveals that they have similar accuracy. Regression analysis and classification tree models in other direct comparisons have also been noted to have comparable performance.^15,16 However, each approach has its advantages and disadvantages. Logistic regression is often preferred when the objective is to obtain explicit measures for statistical inference and to identify the magnitude of the association between each variable and the outcome.¹⁶ However, estimation of the probability of the outcome for each patient with the logistic regression equation is cumbersome and not easily applicable in the NICU. Therefore, we simplified the process by developing a scoring system using odds ratios. The scoring systems provide a numeric value that indicates the magnitude of risk (with a higher score indicating a higher risk). The CART model has fewer variables and does not require any calculations; therefore, it may be simpler to use for clinicians and ancillary staff members. However, CART models provide a qualitative answer (likely or not likely to have a poor outcome) and not a quantitative result, and the magnitude of an individual's risk is not as simple to determine as with the scoring system. Treatment with hypothermia did not lead to significant differences in the correct classification rates for the scoring systems or the CART models, possibly because of limitations in sample size.

It is of interest to compare the variables selected for the scoring systems and the CART analysis. Major contributors to the CART model for death/disability included cord pH, spontaneous activity, and base deficit of the first postnatal blood gas analysis, and major contributors to the scoring systems included posture, spontaneous activity, base deficit of the first postnatal blood gas analysis, and 5-minute Apgar score. The CART models for death had only base deficit of the first postnatal blood gas analysis and cord pH, but major contributors to the regression model included not only the base deficit but also distal flexion/decerebrate posture and absent suck.

One of the useful observations from this study is that not all of the components of the early neurologic examination (eg, level of consciousness, tone, Moro reflex, pupils, heart rate, and respiration) contributed to determination of the prognosis. Abnormal posture, absent spontaneous activity, and absent suck were the only components that contributed significantly to prognosis. Base deficit in the first postnatal arterial blood gas analysis was a variable that was important in both the scoring systems and CART models, and cord pH was also important in the CART model. These results are consistent with previous studies that indicated that the severity of neurologic deficits is related to the degree and duration of metabolic acidosis.^17–19 In the CART model for death/disability, infants with a cord pH of >6.7, normal/decreased spontaneous activity, and a base deficit of 18.5 mmol/L had better outcomes if the PCO₂ of the cord blood gas was >87 mm Hg, compared with 87 mm Hg, which suggests that respiratory acidosis may be associated with better outcomes in the absence of severe metabolic acidosis. This finding is consistent with work by Vannucci et al²⁰ that showed that mild hypercapnia was protective in immature rats with cerebral hypoxia-ischemia.

Variables considered as risk factors for HIE, such as antepartum hemorrhage and chronic hypertension/pregnancy-induced hypertension, were associated with reduced risks of death and/or morbidity in infants who actually developed HIE. We speculate that this paradoxical phenomenon may be attributable to increased clinical vigilance or fetal adaptation. Ischemic preconditioning occurs in neonatal animals,²¹ and intermittent episodes of cerebral ischemia in fetal sheep damage the striatum without injuring the cortex as much.²²

The scoring systems, both for death/disability and for death, identified score ranges of infants who are most likely to benefit from hypothermia. Because of the relatively small sample size and low power, the results of these secondary analyses should be considered hypothesis-generating, rather than a basis for clinical practice. In addition, the relationship of the scores to other clinical outcomes (eg, mild disability) and the effects of hypothermia on those other outcomes have not been determined. It is also possible that some survivors who did not have moderate/severe disability at follow-up assessments and who had low scores on our scoring systems might have disabilities that become apparent only at school age and potentially could have been attenuated by hypothermia. Very few infants had scores high enough that benefit from hypothermia was unlikely. Therefore, our scoring systems are not suitable as a tool for excluding infants from hypothermia.

	CONCLUSIONS

TOP ABSTRACT METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 
Our scoring systems and CART models help in determination of the prognosis of infants with HIE and may be useful in discussions of outcomes with parents and in risk-stratification of infants in future clinical trials. It is important to recognize that these models may be less accurate for a future cohort or a different population, and it is necessary to avoid self-fulfilling prophecies, such as death attributable to withdrawal of support for infants for whom a poor outcome is predicted. It is known that obstetricians and pediatricians who underestimate the possibility of survival of a neonate are less likely to use life-sustaining therapies such as resuscitation, mechanical ventilation, inotropic support, or other standard therapies.²³ With advances in perinatal care or when infants with different characteristics are considered, these algorithms or the interpretations of the scores will need modification to maintain their applicability.

	ACKNOWLEDGMENTS

 
This work was supported by grants from the NICHD and the Department of Health and Human Services (grants U10 HD21385, U10 HD40689, U10 HD27871, U10 HD21373, U01 HD36790, U10 HD40498, U10 HD40461, U10 HD34216, U10 HD21397, U10 HD27904, U10 HD40492, U10 HD27856, U10 HD40521, U10 HD27853, U10 HD27880, and U10 HD27851) and from the National Institutes of Health (grants GCRC M01 RR 08084, M01 RR 00125, M01 RR 00750, M01 RR 00070, M01 RR 0039–43, M01 RR 00039, and 5 M01 RR00044).

Members of the NICHD Neonatal Research Network are as follows: Hypothermia Study Group: Case Western Reserve University, Rainbow Children's Hospital: A.A. Fanaroff, M.C. Walsh, N. Newman, D. Wilson-Costello, B. Siner; Brown University, Women and Infants Hospital: W. Oh, A. Hensman, B. Vohr, L. Noel; Duke University: C.M. Cotten, K. Auten, R. Goldstein, M. Lohmeyer; Emory University, Grady Memorial Hospital, and Crawford Long Hospital: B.J. Stoll, L. Jain, E. Hale; Indiana University, Riley Hospital for Children, and Methodist Hospital: J.A. Lemons, D.D. Appel, L. Miller, A. Dusick, L. Richard; Stanford University: D.K. Stevenson, K. Van Meurs, M.B. Ball, S.R. Hintz; University of Alabama at Birmingham, University Hospital: W.A. Carlo, M. Collins, S. Cosby, M. Peralta-Carcelen, V. Phillips; University of Cincinnati, University Hospital, and Cincinnati Children's Hospital Medical Center: E.F. Donovan, C. Grisby, B. Alexander, J. Shively, H. Mincey, J. Steichen, T. Gratton; University of California, San Diego, University of California, San Diego, Medical Center, and Sharp Mary Birch Hospital for Women: N.N. Finer, D. Kaegi, C. Henderson, W. Rich, K. Arnell, Y.E. Vaucher, M. Fuller; University of Miami: S. Duara, R. Everett, C.R. Bauer; University of Rochester, Golisano Children's Hospital at Strong: R. Guillet, L. Reubens, G. Myers, D. Hust; University of Texas Southwestern Medical Center at Dallas, Parkland Hospital: A.R. Laptook, S. Madison, G. Hensley, N. Miller, R. Heyne, S. Broyles, J. Hickman; University of Texas-Houston, Memorial Hermann Children's Hospital: J.E. Tyson, G. McDavid, E.G. Akpa, C.Y. Franco, P.A. Cluff, A.E. Lis, B.H. Morris, P.J. Bradt; Wayne State University, Hutzel Women's Hospital, and Children's Hospital of Michigan: S. Shankaran, R. Bara, G. Muran, Y. Johnson, D. Kennedy; Yale University, New Haven Children's Hospital: R.A. Ehrenkranz, P. Gettner, E. Romano; NICHD Neonatal Research Steering Committee: Brown University: W. Oh; Case Western University: M.C. Walsh; Duke University: R.N. Goldberg; Emory University: B.J. Stoll; Indiana University: J.A. Lemons; Stanford University: D.K. Stevenson; University of Alabama at Birmingham: W.A. Carlo; University of Cincinnati: E.F. Donovan; University of California, San Diego: N.N. Finer; University of Miami: S. Duara; University of Rochester: D.L. Phelps; University of Texas-Dallas: A.R. Laptook; University of Texas-Houston: J.E. Tyson; Wake Forest University: T.M. O'Shea; Wayne State University: S. Shankaran; Yale University: R.A. Ehrenkranz; University of Cincinnati: A. Jobe, chair; Data Coordinating Center: RTI International: W.K. Poole, B. Hastings, C.M. Petrie; NICHD: R.D. Higgins, L.L. Wright, E. McClure; Data and Safety Monitoring Committee: Children's National Medical Center: G. Avery; Columbia University: M. D'Alton; RTI International: W.K. Poole (ex officio); University of Virginia: J.C. Fletcher (deceased); University of Washington: C.A. Gleason; University of Pittsburgh: C. Redmond.

	FOOTNOTES

 
Accepted Jul 26, 2006.

Address correspondence to Namasivayam Ambalavanan, MD, University of Alabama at Birmingham, 525 New Hillman Building, 619 South 20th St, Birmingham, AL 35249. E-mail: ambal{at}uab.edu

The authors have indicated they have no financial relationships relevant to this article to disclose.

	REFERENCES

TOP ABSTRACT METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

 

1. Lawn J, Shibuya K, Stein C. No cry at birth: global estimates of intrapartum stillbirths and intrapartum-related neonatal deaths. Bull World Health Organ. 2005;83 :409 –417[ISI][Medline]
2. Bryce J, Boschi-Pinto C, Shibuya K, Black RE. WHO Child Health Epidemiology Reference Group: WHO estimates of the causes of death in children. Lancet. 2005;365 :1147 –1152[CrossRef][ISI][Medline]
3. Shankaran S, Laptook AR, Ehrenkranz RA, et al. Whole-body hypothermia for neonates with hypoxic-ischemic encephalopathy. N Engl J Med. 2005;353 :1574 –1584[Abstract/Free Full Text]
4. Gluckman PD, Wyatt JS, Azzopardi D, et al. Selective head cooling with mild systemic hypothermia after neonatal encephalopathy: multicentre randomised trial. Lancet. 2005;365 :663 –670[ISI][Medline]
5. Sarnat HB, Sarnat MS. Neonatal encephalopathy following fetal distress: a clinical and electroencephalographic study. Arch Neurol. 1976;33 :696 –705[Abstract]
6. Robertson CM, Finer NN, Grace MG. School performance of survivors of neonatal encephalopathy associated with birth asphyxia at term. J Pediatr. 1989;114 :753 –760[CrossRef][ISI][Medline]
7. Levene MI, Sands C, Grindulis H, Moore JR. Comparison of two methods of predicting outcome in perinatal asphyxia. Lancet. 1986;1 :67 –69[ISI][Medline]
8. Gunn AJ, Gunn TR, Gunning MI, Williams CE, Gluckman PD. Neuroprotection with prolonged head cooling started before postischemic seizures in fetal sheep. Pediatrics. 1998;102 :1098 –1106[Abstract/Free Full Text]
9. Gunn AJ, Gunn TR, de Haan HH, Williams CE, Gluckman PD. Dramatic neuronal rescue with prolonged selective head cooling after ischemia in fetal lambs. J Clin Invest. 1997;99 :248 –256[ISI][Medline]
10. Gunn AJ, Battin M, Gluckman PD, Gunn TR, Bennet L. Therapeutic hypothermia: from lab to NICU. J Perinat Med. 2005;33 :340 –346[CrossRef][ISI][Medline]
11. Vohr BR, Wright LL, Dusick AM, et al. Neurodevelopmental and functional outcomes of extremely low birth weight infants in the National Institute of Child Health and Human Development Neonatal Research Network, 1993–1994. Pediatrics. 2000;105 :1216 –1226[Abstract/Free Full Text]
12. Palisano R, Rossenbaum P, Walter S, Russell D, Wood E, Galuppi B. Development and reliability of a system to classify gross motor function in children with cerebral palsy. Dev Med Child Neurol. 1997;39 :214 –223[ISI][Medline]
13. Bayley N. Bayley Scales of Infant Development II. San Antonio, TX: Psychological Corp; 1993
14. Brieman L, Friedman JH, Olshen RA, Stone CJ. Classification and Regression Trees. Belmont, CA: Wadsworth; 1984
15. Germanson TP, Lanzino G, Kongable GL, Torner JC, Kassell NF. Risk classification after aneurysmal subarachnoid hemorrhage. Surg Neurol. 1998;49 :155 –163[CrossRef][ISI][Medline]
16. Werneck GL, de Carvalho DM, Barroso DE, Cook EF, Walker AM. Classification trees and logistic regression applied to prognostic studies: a comparison using meningococcal disease as an example. J Trop Pediatr. 1999;45 :248 –251[Abstract/Free Full Text]
17. Low JA, Galbraith RS, Muir DW, Killen HL, Pater EA, Karchmar EJ. Factors associated with motor and cognitive deficits in children after intrapartum fetal hypoxia. Am J Obstet Gynecol. 1984;148 :533 –539[ISI][Medline]
18. Talati AJ, Yang W, Yolton K, Korones SB, Bada HS. Combination of early perinatal factors to identify near-term and term neonates for neuroprotection. J Perinatol. 2005;25 :245 –250[CrossRef][Medline]
19. Toh VC. Early predictors of adverse outcome in term infants with post-asphyxial hypoxic ischaemic encephalopathy. Acta Paediatr. 2000;89 :343 –347[CrossRef][ISI][Medline]
20. Vannucci RC, Towfighi J, Heitjan DF, Brucklacher RM. Carbon dioxide protects the perinatal brain from hypoxic-ischemic damage: an experimental study in the immature rat. Pediatrics. 1995;95 :868 –874[Abstract/Free Full Text]
21. Ota A, Ikeda T, Abe K, et al. Hypoxic-ischemic tolerance phenomenon observed in neonatal rat brain. Am J Obstet Gynecol. 1998;179 :1075 –1078[CrossRef][ISI][Medline]
22. Mallard EC, Williams CE, Gunn AJ, Gunning MI, Gluckman PD. Frequent episodes of brief ischemia sensitize the fetal sheep brain to neuronal loss and induce striatal injury. Pediatr Res. 1993;33 :61 –65[ISI][Medline]
23. Morse SB, Haywood JL, Goldenberg RL, Bronstein J, Nelson KG, Carlo WA. Estimation of neonatal outcome and perinatal therapy use. Pediatrics. 2000;105 :1046 –1050[Abstract/Free Full Text]

This Article Abstract Full Text (PDF) P3Rs: Submit a response Alert me when this article is cited Alert me when P3Rs are posted Alert me if a correction is posted Citation Map Services E-mail this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Add to My File Cabinet Download to citation manager Citing Articles Citing Articles via CrossRef Citing Articles via ISI Web of Science (7) Citing Articles via Google Scholar Google Scholar Articles by Ambalavanan, N. Search for Related Content PubMed PubMed Citation Articles by Ambalavanan, N. Related Collections Premature & Newborn

	Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2006 American Academy of Pediatrics. All rights reserved.