PEDIATRICS Vol. 107 No. 3 March 2001, pp. 588-589

COMMENTARY:
Predicting the Future for Term Infants Experiencing an Acute Neonatal Encephalopathy: Electroencephalogram, Magnetic Resonance Imaging, or Crystal Ball?

In this month's issue, the article entitled "Combined Use of Electroencephalogram and Magnetic Resonance Imaging in Full-Term Neonates With Acute Encephalopathy" by Biagioni et al¹ appears. This report describes the authors' extensive clinical experience evaluating infants presenting at (and shortly before) birth with the ominous quartet of fetal heart rate abnormalities, low Apgar scores, the need for resuscitation at birth, and acute neurologic abnormalities during the first 24 hours of life. The purpose of this study was to investigate the relationship between electroencephalogram (EEG) abnormalities and brain lesions seen on magnetic resonance imaging (MRI), and to determine their prognostic value in neonates with hypoxemic-ischemic encephalopathy (HIE). The authors state that HIE is the most common cause of permanent brain injury in the full-term newborn infant.

Many previous studies both published, and cited by these authors have suggested that neurologic and developmental outcomes can indeed be predicted in term neonates presenting with acute neonatal encephalopathy. These authors investigate here such predictions using either an early EEG obtained before the third day of life, or an MRI scan obtained later, after the first week and before the first month of life. They differentiate 2-year developmental outcomes as either normal, mildly, moderately, or severely abnormal (or died).

They first assigned their 25 cases with acute encephalopathy into 5 EEG background categories, sequenced in order from completely normal to persistently low voltage recordings (most severe), with 3 grades of discontinuity in between (moderate, severe, and extreme). They discovered that 8 infants with normal outcomes at 2 years had a normal (continuous) EEG background, although 1 of these infants had a dysmature EEG (not matching normal background maturity for advanced gestational age). One other infant with a normal EEG had a moderately abnormal outcome.

They next assigned their population into 8 separate MRI categories, sequenced in order from completely normal to severe basal ganglia, thalamic, and diffuse white matter abnormalities (with 6 grades in between). They found 5 infants had normal scans, one had minimal basal ganglia and thalamic changes, and 3 had moderate white matter abnormalities. Of these 9 participants, 2-year outcomes were normal in 8 and only mildly abnormal in 1. Of those with moderate or severely affected MRI grades (total of 16 participants), 11 had moderately to severely abnormal outcomes at age 2 years and 5 had died.

HIE, as a primary diagnostic entity, is an elusive concept and difficult to confirm clinically in the neonate. For example, in our own studies in extracorporeal membrane oxygenation (ECMO)-treated neonates, maximally low PaO₂ levels before ECMO were not significantly related to an increased risk of cerebral palsy (CP),² suggesting that hypoxemia does not necessarily injure the neonatal brain, a concept previously discussed by Vannucci.³ Furthermore, Grether and Nelson⁴ have reported recently that maternal infection increases the risk of CP in surviving term neonates. Forty-six children with disabling spastic CP were compared with 378 control children surviving to 3 years. Maternal fever and chorioamnionitis were associated with a more than ninefold increase in all cases of CP, and a 19-fold increase in quadriplegics. These were the same infants assigned low Apgar scores, who had hypotension and required neonatal resuscitation, had seizures, and were diagnosed as having HIE. Another recent report emphasized the possibility that a difficult-to-diagnose metabolic error, cytochrome oxidase deficiency, may masquerade as HIE.⁵ Many primary pathogenetic mechanisms other than hypoxia or ischemia can result in acute neonatal encephalopathy. HIE and acute neonatal encephalopathy are not, therefore, synonymous.

Obstetricians have also become reluctant to assign perinatal asphyxia as the cause of neonatal HIE based on nonspecific interpretations of fetal heart rate monitoring. Dellinger et al⁶ review specific monitoring criteria for normal, fetal stress, and fetal distress. Neonates with very low Apgar scores, low cord pH, high PCO₂ and a large metabolic base deficit were far more common with specifically defined signs of fetal distress rather than fetal stress on monitoring strips. We need to demand clear evidence connecting postnatal metabolic and neurologic abnormalities with prenatal or neonatal asphyxial events before assigning a diagnosis of perinatally acquired HIE. Additionally, we know that HIE in the neonate, even when fully supported by the available clinical data, is nonspecific in regard to the primary or proximate cause. HIE is an effect not necessarily a cause.

The authors chose not to interpret their data statistically, which is probably prudent because of the large number of different EEG and MRI categories represented by their coding system, and the relatively small and select population evaluated. However, the positive predictive value of using either an EEG or MRI can be calculated from their data.⁷ The presence of any EEG background abnormality observed early in the hospital course predicted 94% of infants who would go on to a mild to severely abnormal outcome. The presence of any MRI abnormality predicted only 85% of those who would have a mild to severely abnormal outcome (attributable to 3 false-positives). There does not seem to be much improvement in predicting any of the abnormal outcomes using both tests.

However, judging by the color-coding used in their Graph 1, the authors did not necessarily consider any and all MRI abnormalities to be detrimental for long-term outcome. Black shading was used only for those participants with moderate basal ganglia and thalamic involvement, or worse (excluding the 3 false-positives above). Recalculating the positive predictive value for the MRI alone using this new threshold for a significantly abnormal study, the authors' data predict 100% of those who had moderate to severely abnormal outcomes, or who died. The EEG adds little to this predictive value.

Obviously, the answer sought depends on the question asked of the data. Clinically, there is little value in tests that predict developmental outcome in neonates who die. And, most neonates with the best test results had either a normal or only mildly abnormal outcome in the authors' study. Which neonates with acute neonatal encephalopathy are destined to survive with severe neurologic impairment is an important question, however. Then, novel and/or extreme interventions might be justified, or highly specialized and unproven therapy might be reasonably withheld or withdrawn as futile.⁸ To answer this question, different positive predictive values can be calculated from the authors' data in Fig 1 in their article, using the moderate EEG discontinuity category (or worse), and the 2 most severe MRI categories as thresholds. Deleting the patients who died from the sample, the respective positive predictive values for a poor outcome are only 36% for the EEG background, but are again 100% for MRI. However, by the time the MRI is abnormal, the damage is already done, so that the use of novel interventions or the withholding of extreme therapies are largely left out of the question. In the neonate at high risk for brain damage, the questions we ask and when we ask them will determine the treatment and the best method for predicting outcome: EEG, MRI, some other test or technology, or conjecture.

Stephen Baumgart, MD
State University of New York at Stony Brook
Stony Brook, NY 11794-8111

Leonard J. Graziani, MD
Thomas Jefferson University
Jefferson Medical College
Philadelphia, PA 19107

FOOTNOTES

Received for publication Oct 27, 2000; accepted Oct 27, 2000.

Address correspondence to Stephen Baumgart, MD, SUNY at Stony Brook School of Medicine, Department of Pediatrics, Health Sciences Center-T11-060, Stony Brook, NY 11794-8111. E-mail: madmaxmd{at}aol.com

ABBREVIATIONS

EEG, electroencephalogram; MRI, magnetic resonance imaging; HIE, hypoxemic-ischemic encephalopathy; ECMO, extracorporeal membrane oxygenation; CP, cerebral palsy.

REFERENCES

1. Biagioni E, Mercuri E, Rutherford MA, Combined use of electroencephalogram and magnetic resonance imaging in full-term neonates with acute encephalopathy. Pediatrics 2001; 107:000-000
2. Graziani LJ, Baumgart S, Desai S, . Clinical antecedents of neurologic and audiologic abnormalities in survivors of neonatal ECMO. J Child Neurol 1997; 12:415-422 [Abstract/Free Full Text]
3. Vannucci RC. Heterogeneity of hypoxic-ischemic thresholds in experimental animals. In: Lou HC, Griesen G, Larson JF, eds. Brain Lesions in the Newborn. Copenhagen, Denmark: Munksgaard; 1994
4. Grether JK, Nelson KB Maternal infection and cerebral palsy in infants of normal birth weight. JAMA 1997; 278:207-211 [Abstract/Free Full Text]
5. Willis TA, Davidson J, George R, Cytochrome oxidase deficiency presenting as birth asphyxia. Dev Med Child Neurol 2000; 42:414-417 [CrossRef][Medline]
6. Dellinger EH, Boehm FH, Crane MM Electronic fetal heart rate monitoring: early neonatal outcomes associated with normal rate, fetal stress, and fetal distress. Am J Obstet Gynecol 2000; 182:214-220 [CrossRef][Medline]
7. Feinstein AR On the sensitivity, specificity, and discrimination of diagnostic tests. Clin Pharmacol Ther 1975; 17:104-116 [Medline]
8. Graziani LJ, Streletz LJ, Baumgart S, . The predictive value of neonatal EEG's before and during ECMO. J Pediatr 1994; 125:969-975 [CrossRef][Medline]