Original Articles

Prognostic Value of EEG in Neonatal Meningitis: Retrospective Study of 29 Infants R o s e m a r y S. C h e q u e r , M D * , B a r r y R . T h a r p , M D * , D a i n a D r e i m a n e , M D * , Jin S. H a h n , M D t , R o b e r t R. Clancy, MD¢, a n d R o n a l d W. C o e n , M D §

Neonatal meningitis is associated with significant neurologic sequelae. Previous studies from our laboratory and others demonstrated electroencephalography (EEG) to be a useful tool in predicting long-term neurologic outcome in at-risk neonates. We, therefore, retrospectively studied 29 infants with culture-proved neonatal meningitis who died in the neonatal period or survived to follow-up at a mean of 34.4 months. Seventy-five EEGs were obtained during the acute phase of infection; the degree of EEG background abnormality proved to be an accurate predictor of outcome. Infants who had normal or mildly abnormal backgrounds had normal outcomes, whereas those with markedly abnormal EEGs died or manifested severe neurologic sequelae at follow-up. When the EEG was considered with the presence or absence of seizures and the level of consciousness, an accurate prediction of neurologic outcome was obtained in 27 infants (93%). Although the EEG patterns were generally nonspecific, some abnormalities, such as positive rolandic sharp waves, persistent hemispheric or focal voltage attenuation, suggested more specific pathology (i.e., deep white matter necrosis, large-vessel infarction and abscess, respectively). EEG was also valuable for the recognition of subtle and subclinical seizures. Therefore, we conclude that EEG is a valuable tool for predicting the long-term prognoses of infants with neonatal meningitis. Chequer RS, Tharp BR, Dreimane D, Hahn JS, Clancy RR, Coen RW. Prognostic value of EEG in neonatal meningitis: Retrospective study of 29 infants. Pediatr Neurol 1992;8:417-22.

Introduction Newborn infants are quite susceptible to bacterial sepsis and many subsequently develop meningitis. In the United States, it is estimated that sepsis occurs in about 1-8:1,000 live births, of whom 25% will develop meningitis [1]. The

From the *Division of Pediatric Neurology; Stanford University Medical Center; Stanford, California; tDivision of Pediatric Neurology; §Department of Pediatrics, Division of Neonatology; University of California, San Diego; San Diego, California; *Division of Pediatric Neurology; Children's Hospital of Philadelphia; Philadelphia, Pennsylvania.

mortality rate for neonatal bacterial meningitis is high (15-20%) despite recent advances in diagnosis and treatment. Morbidity approaches 50% [1-3]. Many studies have examined the neurologic sequelae of neonatal meningitis and reported a variety of clinical, laboratory, and radiologic variables that predicts poor outcomes [2,4-8]. Only 2 studies of neonatal meningitis included a detailed analysis of electroencephalography (EEG) [9,10] which, in our experience with sick newborns, provides valuable prognostic information and may suggest the nature and extent of underlying cerebral pathology [ 11-15]. In this retrospective study, we present our experience with EEGs obtained on 29 premature and term newborns during the acute phase of culture-proved, primarily bacterial, meningitis. Methods Subject Selection. The subjects were selected retrospectively from 3 university intensive care nurseries: Stanford University Medical Center (SUMC), University of California, San Diego (UCSD), and Children's Hospital of Philadelphia (CHOP). All infants with the discharge diagnosis of meningitis were identified by record review at each nursery: 1980 to 1985 at SUMC, 1981 to 1985 at CHOP, and 1983 to 1989 at UCSD. Additionally, the records of the Division of Infectious Disease at SUMC were reviewed from 1985 to 1988 and all neonates with positive cerebrospinal fluid (CSF) cultures were identified. Infants were included in this study if they met all of the following criteria: (1) Positive CSF culture for bacteria or fungus; (2) At least one EEG obtained in the neonatal period during the acute phase of meningitis; and, (3) Follow-up at least l0 months after discharge or death in the neonatal period. Clinical Data. The following data were extracted from the infant's medical record: estimated gestational age (EGA), birth weight, sex, Apgar scores, other significant perinatal neurologic complications (e.g., asphyxia and intraventricular hemorrhage [IVH]), worst level of consciousness (i.e., normal, lethargy, coma), seizures (i.e., any intermittent clinical behavior identified by the nursery staff or consulting pediatric neurologist as "seizure" or "possible seizure"), the most abnormal CSF values (i.e., highest leukocyte count and protein value, and lowest glucose value), the organism identified by CSF and blood culture, the lowest peripheral leukocyte count and absolute neutrophil count, lowest serum bicarbonate level, results of cranial ultrasounds and computed tomographic (CT) scans, and the details of the infant's most recent ex-

Communications should be addressed to: Dr. Tharp; Blue Bird Clinic; Neurosensory Center; Methodist Hospital; 6501 Fannin; Houston, TX 77030. Received March 5, 1992; accepted July 20, 1992.

Chequer et al: EEG in Neonatal Meningitis 417

amination. The charts were also reviewed for any postmeningitic event that may have contributed to a subsequent neurologic deficit. EEG. EEGs were performed in the intensive care nursery by a technique previously described [11]. All infants had at least one EEG during the acute phase of the illness; many had several tracings in the neonatal period. The background cerebral activity was classified according to criteria previously published as normal or mildly, moderately, or markedly abnormal (Table 1) [131. When the infant had more than one EEG, the most abnormal was selected for statistical analysis. When serial EEGs were analyzed and the infants had more than 2 EEGs in I week, the worst tracing was selected. Electrographic seizures were identified on the basis of morphology and duration as previously described [16]. EEGs were interpreted at SUMC by BRT, at UCSD by JSH and BRT, and at CHOP by RRC without the knowledge of outcome. Outcome. The infants were initially examined in the high-risk neonatal programs at the 3 medical centers. Many then were transferred to general pediatric and pediatric neurology clinics. Outcome was classifted as normal, mildly abnormal (i.e., minimal monoparesis, mild motor, or language delay, IQ > 80-85, controlled seizures, hearing loss, arrested hydrocephalus), severely abnormal (i.e., hemiplegia, quadriplegia, uncontrolled seizures, or IQ < 80), and death. Neurodevelopment screening was performed on 13 of 22 survivors and included Bayley Scales of Infant Development (7 patients), Stanford Binet (2), Wechsler Intelligence Scale for Children (1), Reynell Zinkin Developmental Scale for Visual Impairment (1), Amiel-Tison and Knobloch Developmental Screening Inventory (1), and Gesell Examination (1). Radiologic Procedures. Cranial ultrasounds were obtained in 16 patients and CT scans in 12. The findings were classified as: (1) Normal; (2) Mild to moderate ventricular dilatation or isolated intraventriculax hemorrhage as interpreted by the radiologists; or, (3) Parenchymal pathology (e.g., infarction and periventricular leukomalacia). Statistical Analysis. Patients with normal or mildly abnormal outcomes were combined in one group and compared to the second group with severely abnormal outcomes or death. The relationship between outcome and EEG background was tested using the Mann-WhitneyWilcoxon test [17]. Outcome also was compared to the presence of clinical or electrographic seizures using the 2-tailed Fisher exact test.

Results Twenty-nine infants met the criteria for entry into the study (15 from SUMC, 6 from UCSD, and 8 from CHOP). There were 15 males and 14 females with gestational ages ranging from 26-41 weeks (mean: 35.9 weeks). Fifteen were premature (EGA: < 38 weeks). Birth weights were 630-5,170 gm (mean: 2,413 gm). The organism responsible for meningitis was Group B streptococcus (GBS; 13 patients), Escherichia coli (7), Citrobacter (2), Serratia marcesens (2), and 1 case each of Listeria monocytogenes, Aspergillus, Enterobacter aerogenes, Pleisiomonas shigelloide, and Streptococcus faecalis. Clinical Outcome. Seven infants (24%) died in the neonatal period, 14 (48%) survived with no or mild sequelae, and 8 (28%) exhibited severe sequelae. Their ages at follow-up ranged from 10 months to 7 years (mean: 34.4 months). Six of 13 infants with GBS meningitis, the most common cause of infection in this series, had normal outcomes; 2 died and 5 suffered severe neurologic sequelae. EEGs. Seventy-five tracings were recorded during the neonatal period, 62 of which were included in the statistical analysis. The age of the infant at the time of the first

418

PEDIATRIC NEUROLOGY

Vol. 8 No. 6

Figure 1. (A) CT of an infant (EGA: 32 weeks) with Serratia meningitis at 11 days of age reveals multiple areas of low attenuation, more in the right hemisphere consistent with multiple cvsti~ abscesses and cerebral edema. There is contrast enhancement ~/ the h,ptomeninges. (B) Coronal head ultrasonography qf" the same i~!f~tnt at 12 days of age revealed multifocal cystic lesions around the right lateral ventricle. There are echodensities around the &ff't lateral ventricle suggestive of infarcts or abs~ esse~.

recording ranged from 1-42 days (mean: 8.4 days). Twen, ty-three infants had 2 or more EEGs in the 3 weeks following the onset of infection. The relationship between the EEG background and outcome is demonstrated in Table 2.

~

ROC-A1

~-~-~

~ - - ~ .

f

~

~

-

~

-

~

EKG

Chin EMG

~

~

w~

Figure 2. EEG of the same infant as in Figure 1 at 7 days of age (left) demonstrates an electrographic seizure from the left temporal region. There are frequent, multifocal seizures in other parts of the record. At 9 days of age, EEG (right) discloses frequent, positive rolandic sharp waves.

Infants with normal outcomes and mild sequelae were grouped together as were the EEGs classified as normal or mildly abnormal. Normal or mildly abnormal EEGs were always predictive of normal outcome and markedly abnormal tracings were always predictive of abnormal outcomes (severe sequelae and death; P < .0001). The markedly abnormal EEGs were characterized by various combinations of the following patterns: burst-suppression in the preterm (35-38 weeks) and term infants, excessively discontinuous backgrounds in the premature infants (_< 34 weeks of age), seizures, diffuse, nonreactive low-voltage backgrounds, multifocal spikes and sharp waves, persistent regional or focal attenuation of voltage of all background activity, or an isoelectric background. Major asymmetries were observed in 7 infants; 6 who were studied radiologically had one or more of the following: cerebral infarct, single or multiple abscesses, diffuse periventricular white matter necrosis, cerebral edema, and multiple punctate bihemispheric hemorrhages. A single, moderately abnormal EEG was as likely to be followed by normal outcome as severe sequelae. Two or more moderately abnormal EEGs were somewhat more likely to be associated with an abnormal outcome but the number of infants with this EEG finding was too small to be statistically analyzed. Although the backgrounds were similar in the 9 infants with moderately abnormal EEGs, there were several distinctive EEG features in the 4 poor outcome infants with moderately abnormal EEGs that were not present in the 5 with good outcomes. One infant had focal voltage attenuation occurring in several EEGs corresponding to the area of a fungal abscess; that infant had serious neurologic sequelae. Two infants had moderately abnormal backgrounds with positive rolandic sharp waves (PRS). One also had unusual high-voltage delta and prominent delta brushes in the occipital regions, and had severe necrosis of the occipital white matter from Citro-

bacter infection. The second infant with PRS had a large infarct or abscess in the right thalamus, ipsilateral to the PRS (Figs 1,2). The fourth infant with a moderately abnormal EEG and an abnormal outcome had diffuse, nonspecific abnormalities on serial EEGs including focal seizures in the initial recording. Seizures. Clinical episodes, which were believed to represent seizures, occurred in 20 infants (69%); 13 had EEG documentation of seizures. One infant had only electrographic seizures. Three of 14 infants (21%) with EEG ictal activity had normal outcomes, whereas 6 (43%) had severe neurologic sequelae and 5 (36%) died. The presence of electrographic seizures was highly predictive of an abnormal outcome (death or severe sequelae; P < .01). The interictal EEG backgrounds of the 3 infants with seizures and normal or mildly abnormal outcomes were normal in 1 and moderately abnormal in 2, whereas the backgrounds in the 6 with neurologic sequelae were markedly abnormal in 5 and moderately abnormal in 1. These findings suggest that the background is as good a predictor of long-term outcome as the seizure. Normal outcomes occurred in 6 of 20 infants (30%) with clinical seizures, whereas 14 (70%) suffered neurologic sequelae or died. Cranial Imaging Studies. Nineteen infants had a cranial ultrasound and/or CT. Three infants had normal studies, while 8 demonstrated dilated ventricles or isolated intraventricular hemorrhage and 8 had parenchymal abnormalities. All infants with parenchymal abnormalities had neurologic sequelae, whereas 2 infants with normal CT scans (one normal and one with enlarged ventricles) died during the acute phase of the illness. None of the surviving infants with normal studies, isolated ventricular dilatation, or IVH was abnormal at follow-up. Of the 9 infants with normal or mildly abnormal outcomes who had an echo and/or CT, 7 revealed dilated ventricles or IVH and 2 had normal studies.

Chequer et al: EEG in Neonatal Meningitis 419

Table 1. Criteria for classification of EEG background as moderately or markedly abnormal

Moderately Abnormal

(A) Moderately excessive discontinuity for CA (interburst intervals exceeding approximately twice those considered normal tbr the CA for extended periods of the recording) (B) Moderately excessive interhemispheric asynchronyfor CA (C) Poverty of anticipated backgroundrhythms for CA (D) Focal abnormalities including absence of normal background rhythms and transients (E) Moderate hemispheric asymmetry (F) Persistent low-voltage(generalized reduction of voltage [< 25 ~tV] background activity for all states) (G) Dysmaturebackground(background more typical for a CA at least 2 weeks younger) (H) Excessive sharp waves and/or spikes particularly when multifocal Markedly Abnormal

(A) Markedly excessive discontinuity for CA (interburst intervals typically exceeding 3 times the normal range for extended periods of the recording), despite the preservation of some age-appropriate background patterns, such as brushes and rhythmic temporal theta (B) Markedly excessive interhemispheric asynchrony (C) Marked interhemispheric asymmetry (D) Burst suppression (paroxysmal) (E) Diffusely slow backgroundwith paucity of normal transients for CA (F) Extreme low-voltage(< 5 ~tV for all states) (G) Isoelectric Abbreviation: CA = Conceptionalage

A positive relationship existed between the EEG background and radiologic studies. All 5 infants with normal EEGs had cranial imaging studies that were normal or had mild-to-moderate isolated ventricular dilatation or IVH, whereas 6 of 8 infants with markedly abnormal EEG backgrounds had parenchymal abnormalities; the 2 exceptions w e r e infants who died in an early phase of infection. Moderately abnormal backgrounds w e r e recorded in 6 infants; 4 had normal radiologic studies (as d e f i n e d above) and normal outcomes and 2 abnormal studies and severe neurologic sequelae.

[9] and Lerique-Koechlin [101 reported that the EEG wa~ useful in the assessment of infants with neonatal menin gitis and provided valuable prognostic inli)rmation when recorded during the acute phase of infeclion, They reported that the abnormal EEG patterns are rathe, nonspecific and are observed in other acute neonatal encephalopathies (e.g., hypoxic-ischemic encephalopalhy). Llnlortunately, the study by Watanabe et al. included infants without culture-proved meningitis 19] and that of LeriqueKoechlin failed to mention the method used to diagnose meningitis [10]. The present retrospective study relates the experience of 3 neonatal centers with 29 infants with culture-proved infections who were followed-up for at least 10 months (mean: 34.4 months). We found that EEG was a valuable complement to the clinical and radiologic examinations in the prediction of neurologic outcome. In many patients EEG provided unique information suggesting the nature and location of underlying pathology. Infants with normal or mildly abnormal EEG background activity survived without sequelae, and those with markedly abnormal activity died or were neurologically impaired at follow-up. A single EEG, classified as moderately abnormal, did not predict outcome as well as in premature infants with a variety of encephalopathies (primarily hypoxic-ischemic) [ 12]. This latter study flmnd that sequelae were more likely to occur when the moderate EEG abnormality persisted for several weeks. In our study, only 7 infants had serial EEGs that remained in this category and 3 were normal at follow-up. These findings suggest that persistent, moderate abnormalities in an infant with hypoxic-ischemic encephalopathy may indicate a more ominous prognosis than the same EEG findings in an infant with meningitis. Confirmation of this hypothesis awaits future studies of a larger number of infm~ts; however, there were several other distinctive EEG abnormalities that appeared only in the moderately abnormal tracings of infants with poor outcomes. These abnormalities included PRSs (indicative of deep white matter necrosis from ischemia or direct spread of infection) [18] and persistent hemispheric or focal voltage attenuation, which suggested infarction in the distribution of a large vessel or abscess. Therefore, certain electrographic abnormalities, particularly when present in serial recordings, Table 2. The relationship between EEG background and outcome Outcome Most Abnormal EEG Background

Normal or Mildly Abnormal

Severely Abnormal

Death

Normal or mildly abnormal

9

()

()

Moderately abnormal

5

3

t

Markedly abnormal

0

5

6

Discussion

Most studies of neonatal meningitis have not examined the value of the EEG as a prognostic tool. Watanabe et al.

420 PEDIATRICNEUROLOGY Vol. 8 No. 6

suggest the presence of pathology [15,19] and unfavorable outcome. Seizures occurred frequently in these infants with meningitis. Clinical seizures were noted in 69% and electrographic seizures in 48%. Neurologic sequelae or death were common in these infants (79% of infants with EEG seizures and 70% of those with clinical events). The severity of the abnormality of the EEG background, however, appeared to be a better predictor of long-term outcome than the seizures which has been the experience of other investigators who studied large series of infants with seizures of multiple etiologies [20,21]. This point is best illustrated by the one infant with electrographic seizures and normal or mildly abnormal EEG background who was normal at follow-up. Every infant with clinical seizures and markedly abnormal EEG backgrounds had a poor outcome. Moderately abnormal background abnormalities, as was the case with the group as a whole, were less helpful prognostically; 4 of 7 infants with clinical seizures and moderately abnormal EEGs were normal at follow-up. The 3 remaining infants had abnormal cranial imaging studies or were comatose. Most studies of meningitis found that infants with seizures do poorly [2,22,23]. Conversely, Edwards et al. reported that seizures were not significantly associated with abnormal long-term prognoses in a large group of infants with GBS meningitis [4]. This discrepancy may be explained by the method used to diagnose seizures. Recent studies of neonatal seizures using continuous video/EEG monitoring have illustrated that many types of unusual episodic behavior in sick newborns are not accompanied by ictal EEG patterns and are probably not true epileptic seizures [24]; therefore, reliance on only clinical observations may lead to overdiagnosis of seizures and affect outcome results. Conversely, some infants with severe encephalopathy from meningitis may have subclinical seizures that are recognized only by EEG. In a recent study of 41 neonates with seizures verified by EEG (including 7 with meningitis), Clancy et al. reported that 79% of the electrographic seizures were not accompanied by obvious changes in clinical behavior [16]; therefore, EEG is an important tool for the accurate classification of episodic behavior and the recognition of subtle and subclinical seizures. Although the optimal timing of EEG was not specifically studied in this retrospective analysis, it is our recommendation, based on experience with critically ill neonates in other clinical situations, that a recording be obtained during the first week of meningitis when there is a suspicion of seizures as well as during periods of clinical deterioration. All infants with parenchymal abnormalities on CT or cranial ultrasound had poor prognoses and abnormal EEGs. Two infants with moderately abnormal EEGs and abnormal imaging studies did poorly, whereas 4 other infants with moderate EEG abnormalities and normal radiologic studies or isolated ventricular dilatation were normal.

These findings emphasize the prognostic value of performing both procedures, particularly in noncomatose infants. An abnormal EEG may suggest that CT or MRI is indicated, particularly when the EEG is markedly abnormal or moderately abnormal with persistent asymmetry or focal abnormality consistent with infarction or other focal pathology. CT is a useful screening procedure for cerebral infarcts and diffuse cortical abnormalities, but is a rather poor indicator of periventricular leukomalacia, a common complication of meningitis, particularly in the premature infant. The presence of PRSs, which suggests deep white matter lesions, should lead to serial cranial ultrasound studies because radiologically visible lesions may lag behind the appearance of EEG sharp waves.

We thank Lincoln Moses, PhD and Maria Limjoco for their kindness in providing assistance during the statistical analysis of our data, and Raul Bejar, MD for providing the figures.

References

[1] Klein J, Feigin R, McCracken G. Diagnosis and management of meningitis. Pediatrics 1986;78(suppl):959-82. [2] Wald E, Bergman I, Taylor H, Chiponis D, Porter C, Kubek K. Long-term outcome of group B streptococcal meningitis. Pediatrics 1986;77:217-21. [3] Mulder C, Zanen H. A study of 280 cases of neonatal meningitis in The Netherlands. J Infect 1984;9:177-84. [4] Edwards M, Rench M, Haffar A, Murphy M, Desmond M, Baker C. Long-term sequelae of group B streptococcal meningitis in infants. J Pediatr 1985;106:717-22. [5] Chin K, Fitzhardinge P. Sequelae of early-onset group B hemolytic streptococcal neonatal meningitis. J Pediatr 1985;106:819-22. [6] Horn K, Zimmerman R, Knostman J, Meyer W. Neurological sequelae of group B streptococcal neonatal infection. Pediatrics 1974; 53:501-4. [7] Valmari P, M/ikel~i M, Kataja M, Peltola H. Multivariate prognostication in bacterial meningitis of childhood. Scand J Infect Dis 1987; 19:28-34. [8] Pomeroy S, Holmes S, Dodge P, Feigin R. Seizures and other neurologic sequelae of bacterial meningitis in children. N Engl J Med 1990;323:1651-7. [9] Watanabe K, Hara K, Hakamada S, Kuroyanagi M, Kuno K, Aso K. The prognostic value of EEG in neonatal meningitis. Clin Electroencephalogr 1983;14:67-77. [10] Lerique-Koechlin A. EEG in neonatal meningitis and cerebral hemorrhages. In: Rrmond A, ed. Handbook of electroencephalography and clinical neurophysiology. Amsterdam: Elsevier, 1972;15B:53-61. [11] Tharp B. Neonatal and pediatric electroencephalography. In: Aminoff MJ, ed. Electrodiagnosis in clinical neurology. New York: Churchill Livingstone, 1986;77-124. [12] Tharp B, Scher M, Clancy R. Serial EEGs in normal and abnormal infants with birth weights less than 1,200 grams- A prospective study with long-term follow-up. Neuropediatrics 1989;20:64-72. [13] Clancy R, Tharp B, Enzman D. EEG in premature infants with intraventricular hemorrhage. Neurology 1984;34:583-90. [14] Monod N, Pajot N, Guidasci S. The neonatal EEG: Statistical studies and prognostic value in full-term and preterm babies. Electroencephalogr Clin Neurophysiol 1972;32:529-44. [15] Aso K, Scher M, Barmada M. Neonatal electroencephalography and neuropathology. J Clin Neurophysiol 1989;6:103-23. [16] Clancy R, Legido A, Lewis D. Occult neonatal seizures. Epilepsia 1988;29:256-61.

Chequer et al: EEG in Neonatal Meningitis 421

[17] Moses L. Think and explain with statistics. Reading: AddisonWesley, 1986;184-6. [18[ Novotny E Jr, Tharp B, Coen R, Bejar R, Enzmann D. Vaucher Y. Positive rolandic sharp waves in the EEG of the premature infant. Neurology 1987;37:1481-6. [19] Scher M, Tharp B, Sylvestri L. Significance of focal abnormalities in neonatal electroencephalograms: Radiological correlations and outcome. Ann Neurol 1982; 12:217A. [20] Rose A, Lombroso C. Neonatal seizure states: A study of clinical, pathologic, and electroencephalographic features in 137 full term babies with a long-term follow-up. Pediatrics 1970;45:404-25.

422

PEDIATRIC NEUROLOGY

Vol. 8 No. 6

[21] Legldo A, Clancy RR, Berman PH. Neurologic oulcomc alter electroencephalographically proven neonatal seizures Pediatrics 1991: 88:583-96. [22] Pike M, Wong P, Bencivenga R. ei al. Electnq~hy~iologic studies, computed tomography, and neurologic outcome in acnit" bacterial meningitis. J Pediatr 1990;116:702-6. [23] Laxer R, Marks M. Pneumococcal meningiti~ in children. Am J Dis Child 1977; 131:850-3. [24] Mizrahi E, Kellaway P. Characterization and classification of neonatal seizures. Neurology 1987;37:1837-44.