Duration of rhythmic EEG patterns in neonates: new evidence for clinical and prognostic significance of brief rhythmic discharges

Clinical Neurophysiology 111 (2000) 1646±1653

www.elsevier.com/locate/clinph

Duration of rhythmic EEG patterns in neonates: new evidence for clinical and prognostic signi®cance of brief rhythmic discharges AndreÂa J. Oliveira a,b, Magda L. Nunes a,*, LuÂcia M. Haertel a, Fernando M. Reis b, Jaderson C. da Costa a,b a

Division of Neurology, Department of Internal Medicine, SaÄo Lucas University Hospital, PontifõÂcia Universidade CatoÂlica do Rio Grande do Sul, Porto Alegre, Brazil b Department of Physiology, ICBS, Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil Accepted 21 June 2000

Abstract Objective: This study aimed at identifying the characteristics ± especially the duration ± of rhythmic discharges in neonatal EEG, and their association with clinical neonatal problems. Speci®cally, we aimed at testing the diagnostic and prognostic validity of using 10 s as minimal duration for de®ning electroencephalographic seizures. Design and methods: The polysomnographies of 340 neonates were reviewed, and episodes of rhythmic discharges were identi®ed, analyzed, and quanti®ed. The study sample was divided into 3 groups: one in which the maximal duration of rhythmic discharges was shorter than 10 s (brief rhythmic discharges, BRD), a second one in which there were rhythmic discharges longer than 10 s (long rhythmic discharges, LRD), and ®nally a group in which no rhythmic discharge was found (No-RD). These 3 groups of subjects were compared for the baseline and outcome clinical data. Results: From the 340 neonates studied, 210 did not present any form of rhythmic discharge, 67 (19.7%) had only BRD episodes, and 63 (18.5%) had at least one LRD episode. Prevalence of rhythmic discharges was low among healthy full term newborns, and was signi®cantly higher among preterm and high-risk newborns. Electrophysiological characteristics of rhythmic discharges did not differ between healthy neonates and high-risk ones. Accompanying clinical manifestations were present in 26.3% of the LRD group, but also in 15.9% of the BRD group. The presence of BRD was signi®cantly associated with leukomalacia and with hypoglycemia in the cross-sectional analysis of baseline data, and with an increased risk for abnormal neurodevelopmental outcome after a mean follow-up period of 47 months (adjusted relative risk ˆ 4.90, P , 0:01). Conclusions: The present data demonstrate an association between BRD and clinical history of hypoxic-ischemic encephalopathy, especially when complicated by leukomalacia, and also with a prognosis of increased risk for abnormal neurodevelopmental outcome. The clinical and prognostic signi®cance of isolated BRD justi®es the need to include these brief episodes in future studies of neonatal seizures. q 2000 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Newborn; Neonatal EEG; Neonatal seizures; Asphyxia; Outcome

1. Introduction In the last decades, much has been discussed about the dif®culty in de®ning neonatal seizures (Lombroso, 1983, 1996; Mizrahi, 1989; Scher and Painter, 1989; Volpe, 1989; Shewmon, 1990). A fairly reasonable consensus has been reached about the importance of the EEG monitoring, and the necessity to take into consideration the electroencephalographic seizures (Nunes et al., 1995; Lombroso, 1996), * Corresponding author. ServicËo de Neurologia, Hospital SaÄo Lucas da PUC-RS, Av. Ipiranga, 6690, Sala 322, 90610-000 Porto Alegre, RS, Brazil. Fax: 155-51-339-4936. E-mail address: [email protected] (M.L. Nunes).

even when there is no associated clinical expression, since the phenomenon of electroclinical dissociation is frequent in the newborn (Hrachovy et al., 1990; Shewmon, 1990; Weiner et al., 1991). However, a valid de®nition of neonatal ictal EEG patterns, especially regarding their duration, has never been established. Most authors adopt an arbitrary requirement of duration of at least 10 s in order to consider an EEG discharge as ictal (Radvanyi-Bouvet et al., 1985; Scher and Morehead, 1986; Clancy and Legido, 1987, 1991; Clancy et al., 1988; Scher and Beggarly, 1989; Scher et al., 1989; Bye and Flanagan, 1995; Ortibus et al., 1996). Thus, most studies concerning ictal EEG patterns take into account

1388-2457/00/$ - see front matter q 2000 Elsevier Science Ireland Ltd. All rights reserved. PII: S13 88-2457(00)0038 0-1

CLINPH 99138

A.J. Oliveira et al. / Clinical Neurophysiology 111 (2000) 1646±1653

only 10 s episodes or longer, shorter episodes being left out. Some authors do not specify the time limits to consider an EEG event as ictal (Lombroso, 1965, 1974, 1981, 1983; Kellaway and Hrachovy, 1983; Hrachovy et al., 1990). Thus, there is no uniformity in the duration criteria adopted in literature, mostly because this arbitrary cut point has never been evaluated, and, thus far, it has never been shown that episodes shorter than 10 s have no clinical significance, neither is it known whether these brief rhythmic patterns are normal. Shewmon (1990) was the ®rst author to identify this new and arti®cially generated problem. He considered that brief rhythmic discharges might have an epileptogenic value, and called them by the acronym BIRD, which stands for `brief ictal rhythmic discharges', but also for `brief interictal rhythmic discharges' (with an intentional ambiguity re¯ecting the actual dif®culty in distinguishing an interictal from an ictal phenomenon, in general, and especially in neonatal EEG). However, he suggested that BIRDS could be distinguished from other `physiological' brief rhythmic discharges by: (1) their occurrence in infants with seizures and/or encephalopathies; (2) an association, in the same tracing, with longer episodes; and (3) an association with clear, unequivocal seizures of similar morphology in the same tracing. It becomes clear that, in these very selected cases, the presence of unequivocal EEG seizures in the same tracing and of clinical diagnosis of neonatal seizures renders the recognition of BIRDS as an epileptogenic pattern somewhat useless for the accuracy of prospective studies. Meanwhile, the problem of the signi®cance of isolated brief rhythmic discharges remains unsolved. In the present study, we aimed at evaluating whether brief rhythmic discharges alone ± unaccompanied by clear long EEG seizures ± may be considered as a `normal', physiological pattern or, alternatively, represent a pathological pattern, with clinical signi®cance and outcome implications.

2. Design and methods 2.1. Patient population and study design The study included 340 neonates born between the 30th and the 40th weeks of gestation, and submitted to polysomnography at the SaÄo Lucas University Hospital between March 1986 and October 1996. From these, 37 were healthy full term newborns (gestational age $38 weeks) and 26 were preterm neonates without any clinical complication, all born consecutively at the same hospital, and submitted to PSG for purely research purposes. The remaining 277 subjects were high-risk full term or preterm newborns from the Neonatal Intensive Care Unit submitted to PSG as part of their standard clinical care. Indications for polysomnography covered a broad spectrum of neurologic diseases and every condition suggestive of elevated risk of neurologic injury.

1647

All polysomnographies were reviewed blindly, with no previous knowledge of the clinical data of the subject, and all episodes of rhythmic discharges were identi®ed and analyzed. Rhythmic discharges were de®ned as runs of repetitive, rhythmic, monomorphic, stereotyped, sinusoidal or sharp contoured waveforms occurring within a frequency band varying from 0.5 to 20 Hz, thus morphologically resembling normal background rhythms of the adult (Lombroso, 1993) (Fig. 1). No minimal duration was required, other than the minimum necessary to identify a rhythmic pattern. Consequently, episodes as brief as 1 s were included, provided there were enough repetitions of the waves to characterize a rhythmic pattern. Known physiological rhythmic patterns such as delta brushes (Watanabe and Iwase, 1972; Lombroso, 1979, 1993; Torres and Anderson, 1985), frontal delta rhythmic activity (FDRA) (Monod et al., 1972; Arfel et al., 1977; Torres and Anderson, 1985; Lombroso, 1993), premature temporal theta (Torres and Anderson, 1985; Hughes et al., 1987; Lombroso, 1993) and midline theta bursts (Hayakawa et al., 1987; Zaret et al., 1991; Lombroso, 1993) were identi®ed, and were not included in the de®nition of rhythmic discharges. The frequency band, duration and localization of each rhythmic discharge, and the presence of a clinical counterpart to it were recorded. Clinical data were obtained from hospital charts, and included: gestational age (GA); post-conceptional age (CA, de®ned as GA plus the legal age at the moment of the polysomnographic study); Apgar scores at the 1st and the 5th min; any neonatal pathology or complication identi®ed during hospital stay; neuroradiological data whenever available; and general physical and neurologic examinations. These data allowed a cross-sectional analysis of the prevalence, characteristics and clinical signi®cance of rhythmic discharges in the study sample of healthy and high-risk newborns. In order to proceed to a longitudinal cohort evaluation of the prognostic implications of this EEG pattern, the families of the babies involved in the ®rst part of the study were then contacted, the general health status of their child was recorded, and they were asked to bring the child for an interview with the neurologist. A detailed clinical history was then obtained, followed by a complete neurologic examination including assessment of cognitive functions according to patient's age. As a complementary diagnostic tool, the children under 6 years (75.9% of the population studied at follow-up) were also submitted to a quantitative neurodevelopmental assessment using the Denver Developmental Scale. All examinations were carried out by the same physician, without contemporary knowledge of the neonatal PSG data. The main outcome end point was abnormal neurodevelopment. Secondary end points were death from any cause and post-neonatal epilepsy. In healthy neonates, PSG was performed exclusively for research purposes with permission from the institutional

1648

A.J. Oliveira et al. / Clinical Neurophysiology 111 (2000) 1646±1653

review board; parents provided post-informed consent, which was registered in the charts. In high risk newborns, PSGs were part of their standard investigation and were prescribed by the attending physician. The study was authorized by the local Ethics Committee.

2.2. PSG procedures Polysomnographies were recorded in the neurophysiology laboratory, in warmed cribs, usually between 12:00 and 15:00 h, in the interval between two meals. The regis-

Fig. 1. (a) This epoch shows an example of brief rhythmic discharge in the delta band (2±2.5 Hz), lasting 5 s, focally localized on the left central region (electrode C3), without concurrent clinical manifestation. The tracing is characterized by a low voltage background, and frequent episodes of BRD, independently registered at two different electrodes, and occasionally associated with subtle clinical manifestations (convulsive apnea). The patient was a female newborn, with a history of perinatal asphyxia. (b) Example of BRD lasting 6 s, with an evolving pattern: it begins as a run of beta rhythmic sharp contoured waves, changes into alpha frequency, and ends up as theta, always con®ned to the left temporal region (T3). (c) Illustration of a unifocal BRD evolving from the theta (7 Hz) to the alpha band (13 Hz) at F1, lasting 6 s, associated to a convulsive apnea.

A.J. Oliveira et al. / Clinical Neurophysiology 111 (2000) 1646±1653

tration lasted at least 1 h or a complete sleep cycle. At least 9 scalp electrodes were placed according to the International 10±20 system modi®ed for neonates (Lombroso, 1993; Nunes et al., 1997), with double electrode distance, in the positions F1-2, C3-4, T3-4, O1-2, and Cz. A 13-channel Toshiba, or a 16-channel Berger machine was used to record at least 8 channels of EEG in a standard bipolar montage, and 5 extracerebral monitoring channels, including submental electromyogram, eye movement, electrocardiogram, nasal thermistor, and thoraco-abdominal respiratory movements, with a paper speed of 15 mm/s. 2.3. Data analysis The study sample was divided into 3 groups: one group in which the maximal duration of rhythmic discharges was shorter than 10 s (brief rhythmic discharges, BRD), a second one in which the maximal duration of rhythmic discharges was equal or longer than 10 s (long rhythmic discharges, LRD), and ®nally a group in which no rhythmic discharge was found (No-RD). These 3 groups of subjects were compared for the baseline and outcome clinical data. With the purpose of testing the diagnostic and prognostic validity of the 10 s cut point usually adopted in literature for de®ning electroencephalographic seizures, the LRD group was compared with all the remaining subjects (BRD and No-RD groups merged), simulating what happens in those studies which take into account only episodes lasting longer than 10 s. In order to test the diagnostic and prognostic validity of BRD, this group was compared to the group without rhythmic discharges. 2.4. Statistical analysis The 3 independent study groups (No-RD, BRD, LRD) were compared for continuous variables (GA, CA, Apgar Scores) using one-way analysis of variance (ANOVA), and for dichotomous variables using chi-square tests with linearity correction. Signi®cance level was set at P , 0:05. Multivariate logistic regression was applied to determine whether BRD or LRD were independently associated with the baseline clinical measures. In the cross-sectional analysis of baseline data, the presence of only BRD (BRD group) was considered as a primary end point, and tests were performed to identify clinical features associated with the presence of this pattern, versus absence of rhythmic patterns (no rhythmic discharges at all). As a second step of the cross-sectional analysis, the end point was the presence of at least one episode of LRD, which was compared to the absence of LRD (no rhythmic discharges or only BRD). Data are expressed as odds-ratio (OR) with 95% con®dence intervals (CI). Outcome data were subjected to univariate and multivariate analysis in order to verify whether BRD or LRD were independent risk factors to unfavorable outcome. Two separate regression models were constructed for multivariate analysis. In the ®rst one, the presence of at least one episode

1649

of LRD was considered as potential risk factor, and tested by confrontation with absence of LRD (only BRD or NoRD). In the second model, the presence of any rhythmic discharge (LRD or only BRD) was considered as a potential risk factor, and therefore tested by confrontation with the absence of rhythmic patterns through a stepwise forward method to adjust for confounding variables. Potentially confounding variables were those baseline clinical problems associated with rhythmic discharges and per se determinants of poor outcome, as determined by univariate analysis. Data are expressed as relative risk (RR) with 95% CI. Statistical power of the cohort was calculated using the package Epiinfo 6.04b (CDC, Atlanta), setting the expected incidence of disease in unexposed groups to that observed in the same population.

3. Results 3.1. Prevalence and electrophysiological characteristics of rhythmic discharges Of the 340 neonates studied, 210 did not show any form of rhythmic discharges, 67 (19.7%) had only BRD episodes, and 63 (18.5%) had at least one episode of LRD. Prevalence of rhythmic discharges was low among healthy full term newborns (10.8%). Rhythmic discharges were signi®cantly more prevalent among healthy preterm newborns (30.8%, P ˆ 0:05, vs. term), mostly due to an increase in LRD prevalence, and among high-risk newborns (term, 44.7%; preterm, 48.5%; P , 0:001 vs. healthy), due to an increase of both BRD and LRD prevalence (Fig. 2). Electrophysiological characteristics of rhythmic discharges (i.e. the main frequency band, the maximal and

Fig. 2. Prevalence of rhythmic discharges among healthy and high-risk neonates born at term or preterm. *P , 0:05 (Chi-square test).

1650

A.J. Oliveira et al. / Clinical Neurophysiology 111 (2000) 1646±1653

minimal duration of the episodes during a tracing, the variability of this duration, and the number of episodes in each tracing) were similar in high-risk and healthy neonates. In both cases, intermediate frequency bands (alpha and theta) were the most prevalent ones (x2 ˆ 46:2; d:f: ˆ 3; P , 0:00001). Median duration of the longest rhythmic discharge in each tracing was 9 s (range 1±3600 s; the latter corresponds to one high-risk neonate in whom there was a continuous focal rhythmic discharge throughout the recording period). The median number of rhythmic discharges along a tracing was 3 (range 1±30). Most rhythmic discharges were unifocal; they were multifocal in only 6 high-risk newborns, and generalized in none. Episodes of rhythmic discharges were found in all electrodes (frontal, central, temporal, occipital), without hemispheric predominance, but rarely with a bilateral synchronous distribution or over the midline. 3.2. Baseline clinical features vs. BRD and LRD Mean gestational age, conceptional age and Apgar scores in the 1st and 5th min did not signi®cantly differ among the 3 study groups (i.e. BRD, LRD, and No-RD groups, Table 1). Table 2 shows the frequency distribution of the most prevalent clinical problems among the 277 high-risk neonates included in the study population. The occurrence of BRD alone was signi®cantly associated with hypoxicischemic encephalopathy (HIE) (OR ˆ 1.94, 95% CI 1.05±3.56), leukomalacia (OR ˆ 3.89, 95% CI 1.14± 13.21), and hypoglycemia (OR ˆ 2.72, 95% CI 1.06± 6.90). After adjustment for concurrent factors, leukomalacia (adjusted OR ˆ 3.94, P , 0:05) and hypoglycemia (adjusted OR ˆ 2.74, P , 0:05) con®rmed their independent association with BRD (Table 3). The presence of LRD was signi®cantly associated with HIE (OR ˆ 1.83, 95% CI 1.00±3.36) and peri and intraventricular hemorrhage (OR ˆ 2.28, 95% CI 1.01±5.14), but only HIE was shown to have an independent association with LRD after multivariate analysis (adjusted OR ˆ 1.86, P , 0:05, Table 3). The prevalence of background EEG abnormalities, such as interhemispheric asymmetry (3.3±4.8%), low voltage background (14.8±20.6%), and burst-suppression pattern (1.4±6.3%) were similar among the 3 study groups. Table 1 Basal characteristics of the 3 study groups (No-RD, BRD and LRD) regarding gestational age (GA), conceptional age (CA) and Apgar scores in the 1st and in the 5th min a Group

N

GA (weeks)

CA (weeks)

Apgar 1 0

Apgar 5 0

No-RD BRD LRD

210 67 63

36.2 ^ 0.3 36.0 ^ 0.4 35.5 ^ 0.5

38.8 ^ 0.3 38.8 ^ 0.4 38.3 ^ 0.5

6.7 ^ 0.3 6.2 ^ 0.5 5.7 ^ 0.4

8.1 ^ 0.2 7.9 ^ 0.3 7.9 ^ 0.2

a

Data are means ^ SEM.

Table 2 Frequency distribution of the most prevalent clinical problems among the 277 high-risk newborns included in the study population a Clinical problems

n

%

Prematurity Hypoxic-ischemic encephalopathy b Neonatal infection c Intra and periventricular hemorrhage Respiratory distress syndrome CNS malformation Hypoglycemia d Leukomalacia Congenital infection Intrinsic metabolic defects

106 95

38.3 34.3

50 35

18.1 12.6

29 26 25 14 12 11

10.5 9.4 9.0 5.1 4.3 4.0

a Note: the sum of percentages exceeds 100 because of superimposition of clinical problems. b De®ned by the presence of clear-cut neurological signs of neonatal encephalopathy with documented birth asphyxia (depressed Apgar scores, cord blood acidosis) and/or neuroradiological evidence of hypoxicischemic lesions. c Infective diseases contracted after birth, comprising bacterial meningitis, encephalitis, and sepsis. d Blood glucose levels under 20 mg/dl.

Ictal clinical manifestations associated with the electroencephalographic rhythmic patterns were not found in any of the 12 healthy neonates who presented rhythmic discharges, but were found in 25 out of 109 (22.9%) highrisk newborns with rhythmic discharges in whom concomitant clinical observation had been documented. Clinical manifestations in association with the electroencephalographic rhythmic discharges were more frequent in the LRD group (15 out of 57 documented cases; 26.3%), but clinical manifestations were also present in 10 out of 63 Table 3 Clinical variables possibly associated with increased prevalence of brief rhythmic discharges (BRD) or long rhythmic discharges (LRD) a Variable

OR

95% CI

Hypoxic-ischemic encephalopathy BRD 1.94 1.05±3.56 LRD 1.83 1.00±3.36 Leukomalacia BRD 3.89 1.14±13.21 LRD 1.25 0.33±4.63 Peri and intraventricular hemorrhage BRD 2.26 0.91±5.57 LRD 2.28 1.01±5.14 Hypoglycemia BRD 2.72 1.06±6.90 LRD 0.89 0.29±2.74 Prematurity BRD 1.58 0.85±2.92 LRD 1.54 0.82±2.93

Adjusted OR

P

1.86

0.104 0.043

3.94 ±

0.029 0.995

± ±

0.144 0.118

2.74 ±

0.036 0.694

± ±

± ±

a OR, odds ratio; CI, con®dence interval. The stepwise multivariate model tested the effect of all variables signi®cantly associated with BRD or LRD.

A.J. Oliveira et al. / Clinical Neurophysiology 111 (2000) 1646±1653

subjects (15.9%) of the BRD group. The most frequent clinical manifestations observed were focal clonic or focal tonic seizures, convulsive apneas, and motor automatisms. It is noteworthy that, from the 25 cases in which rhythmic discharges were correlated with a clinical event, 10 (40%) were cases of exclusive BRD.

Of the 340 neonates included in the cross-sectional study, 113 were reevaluated in follow-up, with ages varying from 6 months to 12 years (mean 3 years and 11 months). In this cohort, the groups with and without RD had comparable baseline characteristics like GA, CA and Apgar scores. The follow-up period was equivalent among the study groups as regards mean, range and dispersion of ages at the time of outcome evaluation (F ˆ 0:86, P ˆ 0:42, ANOVA). Incidence of post-neonatal epilepsy was similar in the 3 groups (Table 4). The mortality rate was higher in the BRD group (3 out of 29, 10.3%) and in the LRD group (two out of 31, 6.5%) than in the No-RD group (one out of 53, 1.9%), but this difference did not reach statistical significance, probably due to the low incidence. The presence of BRD episodes in the neonatal EEG was signi®cantly correlated with an increased incidence of abnormal neurodevelopmental outcome (48.1 vs. 17.0% in the No-RD group; x2 ˆ 7:61, P , 0:01, Table 4). The group with LRD had an incidence of 37.9% of abnormal neurodevelopmental outcome, which was also signi®cantly higher than the incidence in the No-RD group (x2 ˆ 3:83, P , 0:05). The presence of any rhythmic discharge (LRD or BRD) determined an increased risk of abnormal neurodevelopmental outcome (RR ˆ 1.70, 95% CI ˆ 1.24±2.34). This risk increment was independent of the risk imputable to the underlying neonatal pathology, as shown after adjustment for confounding variables in multivariate analysis (adjusted RR ˆ 4.05, P , 0:01; Table 5). The presence of only BRD was suf®cient to increase the risk of unfavorable outcome (adjusted RR ˆ 4.90 vs. No-RD, P , 0:01). If the concept of rhythmic discharges had included only episodes lasting at least 10 s (LRD), this EEG ®nding would no longer be sensitive enough to predict an increased risk of abnormal neurodevelopmental outcome. This would happen because such a de®nition implies in classifying neonates with only BRD Table 4 Incidences of abnormal neurodevelopmental outcome, epilepsy and death in the groups BRD, LRD and No-RD after a mean follow-up of 47 months a Group

N

Abnormal neurodevelopmental outcome

Epilepsy

Death

No-RD BRD LRD

53 29 31

9 (17.0%) 14 (48.3%)** 12 (38.7%)*

8 (15.1%) 5 (17.2%) 7 (22.6%)

1 (1.9%) 3 (10.3%) 2 (6.5%)

*P , 0:05, **P , 0:01 vs. No-RD (chi-square test).

together with those with No-RD at all, with a consequent marked decrease of the RR (RR ˆ 1.38, 95% CI ˆ 0.79± 2.42, Table 5). HIE was also associated with a signi®cant increase of risk of abnormal neurodevelopmental outcome, as expected (adjusted RR ˆ 9.96, P , 0:001). 4. Discussion

3.3. Prognostic value of rhythmic discharges

a

1651

There are many references in the literature to the existence of very brief electroclinical seizures in the neonatal period (Dreyfus-Brisac and Monod, 1972; Lombroso, 1974, 1983; Engel, 1975; Lou and Friis-Hansen, 1979; Aicardi, 1986; Da-Costa et al., 1992). Nevertheless, most studies, particularly those evaluating the prognostic signi®cance of EEG seizures, still adopt the arbitrary limit of 10 s for the de®nition of ictal discharges. Overlooking brief episodes, however, may dramatically affect the results of these studies, since, as we have demonstrated in the present study, brief rhythmic discharges per se are associated with speci®c clinical neonatal problems, and bear an important prognostic implication, being associated with an increased risk for abnormal neurodevelopmental outcome. It must be noted that in as many as 40% of the subjects with rhythmic discharges accompanied by a clinical seizure, the electrographic rhythmic discharge lasted less than 10 s. Thus, using 10 s as a cutoff point, in this case, would mean excluding more than one-third of the tracings with an electroclinical seizure. Whether these brief rhythmic discharges, when unaccompanied by clinical manifestations, constitute an electrographic seizure, or an interictal pattern, or even a nonepileptogenic abnormal EEG pattern, has still to be determined. A clue to this question could be given by the fact that, in 15.9% of the babies with only BRD, this pattern was accompanied by clinical manifestations ± a percentage not signi®cantly different from that of the LRD group (26.3%). The remaining episodes could possibly correspond to electrographic seizures without clear clinical manifestations. The issue of electroclinical dissociation has long been recognized among neonatal electroencephalographers (Weiner et al., 1991; Lombroso, 1993, 1996). Bye and Flanagan (1995) found that 85% of all recorded EEG seizures in neonates with con®rmed electroclinical seizures were unaccompanied by clinical signs. In the same study, however, these authors demonstrated that prolonged recordings allowed a high percentage (88%) of clinical manifestations of EEG seizures to be eventually identi®ed. Another possibility is that isolated BRD without clinical manifestation correspond to an interictal pattern. Here we arrive at a completely open matter: ictal/interictal boundaries are often somewhat blurred in other ages, but this is particularly true in the neonatal period. In all instances, however, brief rhythmic discharges would retain their clinical relevance and their contribution to an increased risk of abnormal neurodevelopmental outcome.

1652

A.J. Oliveira et al. / Clinical Neurophysiology 111 (2000) 1646±1653

Table 5 Increased risk of abnormal neurodevelopmental outcome associated with the presence of rhythmic discharges during neonatal life Risk factor a

LRD (vs. only BRD or No-RD) LRD or BRD (vs. No-RD) b BRD (vs. No-RD) c LRD (vs. No-RD) d Peri and intraventricular hemorrhage Leukomalacia HIE

RR

95% CI

Adjusted RR

P

1.38 1.70 2.84 2.23 3.11 2.56 3.15

0.79±2.42 1.24±2.34 1.41±5.75 1.05±4.76 1.97±4.89 1.46±4.49 1.85±5.38

Not in model 4.05 4.90 2.86 Not in model Not in model 9.96

0.278 ,0.01 ,0.01 ,0.05 0.108 0.225 ,0.01

a LRD vs. only BRD or No-RD simulates the use of 10 s cutoff to de®ne a neonatal seizure; LRD or BRD vs. No-RD ( b) considers the risk associated with any rhythmic discharge including those lasting as little as 1 s. In addition, the speci®c impacts of BRD ( c) and LRD ( d) on the risk of unfavorable outcome were calculated separately. In the multivariate analysis, the variable RD was entered in several rounds of stepwise forward logistic regression with different cutoffs and designs. Relative risks (RR) were adjusted to potential confounding variables (peri and intraventricular hemorrhage, leukomalacia, and hypoxic-ischemic encephalopathy [HIE]). The adjusted RR for HIE refers to model ( b).

The increased risk of abnormal outcome associated with rhythmic discharges in the longitudinal part of our study is independent of the risk already determined by the concurrent factors, as shown after a multivariate analysis. Clearly, postneonatal morbidity is determined by the underlying pathology to which rhythmic discharges are associated ± as suggested by many studies (Lombroso, 1996; Clancy and Legido, 1991; Ortibus et al., 1996), and as our data reinforce by showing an adjusted relative risk of 9.96 for abnormal neurodevelopmental outcome associated with HIE. It is important, however, to understand that the presence of rhythmic discharges ± even of brief duration ± in the neonatal EEG adds a risk to the prognosis of the patient. Two hypotheses could explain the additional risk imposed by the occurrence of rhythmic discharges in the neonatal period. The ®rst hypothesis refers to the question of whether repeated ictal discharges may cause neuronal damage, or further aggravate the original brain damage. Experimental models of neonatal seizures suggest that this may happen, although results are contradictory and there is still a gap between laboratory ®ndings and the clinic (Volpe, 1987; Clancy et al., 1988; Holmes, 1988; Scher and Painter, 1989; Shewmon, 1990; Engel et al., 1992; Lombroso, 1996, 1998). The structural damage might be caused by metabolic mechanisms leading to energy failure, with a consequent cascade of deleterious effects (Wasterlain and Duffy, 1976), eventually resulting in brain damage and subsequent neurodevelopmental de®cits, or, alternatively, by an excessive release of endogenous excitatory amino acids, which would eventually result in an intracellular calcium overload and subsequent cell death (Johnston and McDonald, 1993). An association between both mechanisms is also possible. Anyway, the matter is not closed, even for prolonged seizures. For brief discharges, the hypothesis would seem weaker, although it cannot be discarded. A second, reasonable hypothesis should be considered to explain how the occurrence of brief discharges in the neonatal period determines an increased risk for abnormal outcome. The presence of these discharges, even if not dele-

terious per se, could be an epiphenomenon associated to and indicative of a more severe underlying pathology. Thus, although clinically similar, neonates with and without rhythmic discharges on the EEG could represent different subgroups from the same neonatal problem. Different degrees of brain damage, maybe too subtle to be noticed clinically or on neuroimaging, could determine the difference between the occurrence or not of these activities, and, concomitantly, a worse prognosis. In our results, the fact that LRD episodes failed to show any association with outcome measures, while exclusive BRD did show this kind of association, does not mean that BRD are more `severe', or have a higher impact on perinatal neurological development than LRD. It simply re¯ects an effect of the placement of BRD together with no rhythmic discharges, in contraposition to the group in which there was at least one episode of rhythmic discharge longer than 10 s. This placement mimics what happens when a 10 s cut point is set in most studies of neonatal seizures. The present results show that the setting of a cut point of 10 s for the de®nition of rhythmic discharges clearly decreases the predictive value of neonatal EEG. In conclusion, the present data demonstrate an association between BRD and clinical history of HIE, especially when complicated by leukomalacia, with hypoglycemia, and also with a prognosis of increased risk for abnormal neurodevelopmental outcome. The ®nding of rhythmic discharges in the neonatal EEG should thus alert the physician to the risk of a guarded prognosis. Furthermore, the clinical and prognostic signi®cance of isolated BRD justi®es the need to include these episodes when characterizing neonatal seizures in future studies.

Acknowledgements This work was supported in part by a grant to A.J.O. from CoordenacËaÄo de AperfeicËoamento do Pessoal de Ensino Superior (CAPES, Brazil).

A.J. Oliveira et al. / Clinical Neurophysiology 111 (2000) 1646±1653

References Aicardi J. Neonatal seizures. In: Aicardi J, editor. Epilepsy in children, New York: Raven Press, 1986. pp. 183±204. Arfel G, Leonardon N, Moussalli F. Densite et dynamique des encoches pointues frontales dans le sommeil du nouveau-ne et du nourrisson. Rev EEG Neurophysiol 1977;7:351±360. Bye AME, Flanagan D. Spatial and temporal characteristics of neonatal seizures. Epilepsia 1995;36:1009±1016. Clancy RR, Legido A. The exact ictal and interictal duration of electroencephalographic neonatal seizures. Epilepsia 1987;28:537±541. Clancy RR, Legido A. Postnatal epilepsy after EEG-con®rmed neonatal seizures. Epilepsia 1991;32:69±76. Clancy RR, Legido A, Lewis D. Occult neonatal seizures. Epilepsia 1988;29:256±261. Da-Costa JC, Nunes ML, Nora DB, Lopes DK, Perondi G, Fagundes SC. Apneic seizures in newborns. Pediatr Neurol 1992;8:382. Dreyfus-Brisac C, Monod N. Neonatal status epilepticus. In: ReÂmond A, editor. Handbook of electroencephalography and clinical neurophysiology, Amsterdam: Elsevier, 1972. pp. 38±52. Engel JJ, Wasterlain C, Cavalheiro EA, Heinemann U, Avanzini G. Molecular neurobiology of epilepsy. Epilepsy Res 1992: Suppl 9. Engel RCH. Abnormal electroencephalograms in the neonatal period, Spring®eld, IL: Thomas, 1975. Hayakawa F, Watanabe K, Hakamada S, Kuno K, Aso K. Fz theta/alpha bursts: a transient EEG pattern in healthy newborns. Electroenceph clin Neurophysiol 1987;67:27±31. Holmes GL. Do seizures cause brain damage? Int Pediatr 1988;3:158±164. Hrachovy RA, Mizrahi EM, Kellaway P. Electroencephalography of the newborn. In: Daly DD, Pedley TA, editors. Current practice of clinical electroencephalography, New York: Raven Press, 1990. pp. 201±242. Hughes JR, Fino JJ, Hart LA. Premature temporal theta. Electroenceph clin Neurophysiol 1987;67:7±15. Johnston MV, McDonald JW. Metabolic and pharmacological consequences of seizures. In: Dodson WE, Pollock JM, editors. Pediatric epilepsy, New York: Demos Public, 1993. pp. 27±35. Kellaway P, Hrachovy RA. Status epilepticus in newborns: a perspective on neonatal seizures. In: Delgado-Escueta AV, Wasterlain CG, Treiman DM, Porter RJ, editors. Advances in neurology, vol. 34. Status epilepticus, New York: Raven Press, 1983. pp. 93±99. Lombroso CT. Neonatal seizure states. In: 11th Intl. Congr. Pediatr. Proceed., Tokyo: University of Tokyo Press, 1965. pp. 38±49. Lombroso CT. Seizures in the newborn period. In: Vinken PJ, Bruyn GW, editors. Handbook of clinical neurology, Amsterdam: North-Holland, 1974. pp. 189±218. Lombroso CT. Quanti®ed EEGraphic scales on 10 preterm healthy newborns followed up to 40-43 weeks by serial polygraphic recordings. Electroenceph clin Neurophysiol 1979;46:460±474. Lombroso CT. Normal and abnormal EEGs in full-term neonates. In: Henry C, editor. Current clinical neurophysiology, Amsterdam: Elsevier/ North-Holland, 1981. pp. 83±150. Lombroso CT. Neonatal seizures. In: Browne TR, Feldman RG, editors. Epilepsy diagnosis and management, Boston, MA: Little, Brown & Co, 1983. pp. 297±313. Lombroso CT. Neonatal EEG polygraphy in normal and abnormal

1653

newborns. In: Niedermeyer E, Lopes da Silva F, editors. Electroencephalography: basic principles, clinical applications and related ®elds, Baltimore, MD: Williams & Wilkins, 1993. pp. 803±875. Lombroso CT. Neonatal seizures: historic note and present controversies. Epilepsia 1996;37:5±13. Lombroso CT, Nunes ML, da-Costa JC. Crises convulsivas no receÂmnascido. In: da-Costa JC, Palmini A, Yacubian EMT, Cavalheiro E, editors. Fundamentos neurobioloÂgicos das epilepsias: aspectos clõÂnicos e ciruÂrgicos, SaÄo Paulo: Lemos Editorial, 1998. Lou HC, Friis-Hansen B. Arterial blood pressure elevations during motor activity and epileptic seizures in the newborn. Acta Paediatr Scand 1979;68:803±806. Mizrahi EM. Consensus and controversy in the clinical management of neonatal seizures. Clin Perinatol 1989;16:485±500. Monod N, Pajot N, Guidasci S. The neonatal EEG: statistical studies and prognostic value in full-term and pre-term babies. Electroenceph clin Neurophysiol 1972;32:529±544. Nunes ML, da-Costa JC, Taufer L, da-Silveira CM. Value of EEG in the characterization and prognosis of neurological diseases in premature infants. Arq Neuropsiquiatr 1995;53(3):625±630. Nunes ML, da-Costa JC, Moura-Ribeiro MVL. Polysomnographic quanti®cation of bioelectrical maturation in preterm and full-term newborns at matched conceptional ages. Electroenceph clin Neurophysiol 1997;102:186±191. Ortibus EL, Sum JM, Hahn JS. Predictive value of EEG for outcome and epilepsy following neonatal seizures. Electroenceph clin Neurophysiol 1996;98:175±185. Radvanyi-Bouvet MF, Vallecalle MH, Morel-Kahn F, Relier JP, DreyfusBrisac C. Seizures and electrical discharges in premature infants. Neuropediatrics 1985;16:143±148. Scher MS, Beggarly ME. Clinical signi®cance of focal periodic discharges in neonates. J Child Neurol 1989;4:175±185. Scher MS, Morehead LM. Electrographic seizures in the high risk newborn. Epilepsia 1986;27:609±612. Scher MS, Painter MJ. Controversies concerning neonatal seizures. Pediatr Clin North Am 1989;36:281±310. Scher MS, Painter MJ, Bergman I, Barmada MA, Brunberg J. EEG diagnosis of neonatal seizures: clinical correlations and outcome. Pediatr Neurol 1989;5:17±24. Shewmon DA. What is a neonatal seizure? Problems in de®nition and quanti®cation for investigative and clinical purposes. J Clin Neurophysiol 1990;7:315±368. Torres F, Anderson C. The normal EEG of the human newborn. J Clin Neurophysiol 1985;2:89±103. Volpe JJ. Neurology of the newborn, Philadelphia, PA: W.B. Saunders, 1987. Wasterlain CG, Duffy TE. Status epilepticus in immature rats: protective effects of glucose on survival and brain development. Arch Neurol 1976;33(12):831±837. Watanabe K, Iwase K. Spindle-like fast rhythms in the EEGs of low birthweight infants. Dev Med Clin Neurol 1972;14:373±381. Weiner SP, Painter MJ, Geva D, Guthrie RD, Scher MS. Neonatal seizures: electroclinical dissociation. Pediatr Neurol 1991;7:363±368. Zaret BS, Guterman B, Weig S. Circumscribed midline EEG activity in neurologically normal neonates. Clin Electroenceph 1991;22:13±22.