Article Text

Download PDFPDF

Prognostic value of EEG in very premature newborns
  1. A Le Bihannic1,2,
  2. K Beauvais1,
  3. A Busnel3,
  4. C de Barace2,4,
  5. A Furby1
  1. 1Unité de Neurophysiologie Clinique, Hôpital Yves Le Foll, Saint Brieuc, France
  2. 2Service de Pédiatrie, Hôpital Yves Le Foll, Saint Brieuc, France
  3. 3Service de Néonatalogie, Hôpital Yves Le Foll, Saint Brieuc, France
  4. 4CAMSP, Saint Brieuc, France
  1. Correspondence to Dr A Furby, Unité de Neurophysiologie Clinique, Hôpital Yves Le Foll, 22027 Saint Brieuc, France; alain.furby{at}


Objective To evaluate the prognostic value of EEG regarding the psychomotor outcomes of very premature newborns.

Methods 76 premature infants <30 weeks gestation were enrolled between January 2001 and August 2004. They were examined at 4 and 9 months corrected ages, and at 18 months, 3–4 years and 5–6 years. EEGs performed in the neonatal period were analysed by two neurologists blind to the child's outcome.

Results The mean follow-up was 5.6 years. 25 infants had normal neurological development and all EEGs were normal for 22 of these. 36 others had developmental disabilities (7 motor sequelae and 29 delayed psychomotor development). Of 187 EEGs, 43 were dysmature, 13 disorganised, 2 displayed electrical seizures without clinical manifestations and 15 showed other abnormal features. Dysmaturity was the predominant EEG pattern in newborns with severe or moderate sequelae and was persistent on several EEGs in 12 of these. In contrast, only three infants with normal development had a dysmature pattern on one EEG. All infants with a disorganised pattern had cognitive sequelae, and two had cerebral palsy. The sensitivity of EEG regarding psychomotor outcome was 83.3%, the specificity was 88% and the positive predictive value was 90.9%.

Conclusions Very preterm neonates remain at high risk of neurological sequelae and EEG is a sensitive method for assessing neuromotor and cognitive prognosis. A dysmature pattern was the predominant EEG characteristic in infants who developed severe or moderate impairment. Early postnatal tracing is useful but additional recordings are generally necessary to detect high-risk newborns.

View Full Text

Statistics from

Over the past decades, advances in neonatal intensive care have improved the survival rate of preterm infants1,,3 and in particular very premature newborns of <30 weeks gestational age (GA).3 4 However, an increased incidence of neurological sequelae and developmental disorders is observed in these preterm survivors.5,,7 Consequently, there is a need to identify high-risk newborns as early as possible so they can be provided with specific care. EEG is one of the most useful and non-invasive tools for predicting neurological outcome in preterm and term infants.6 8,,14 However, previous reports have mainly focused on severe motor sequelae, and children were not usually followed past 1 or 2 years after birth.4 8 11 The aim of this study was to correlate EEG findings in very premature infants with their neurological and neuropsychological outcomes beyond 2 years of age.


We retrospectively investigated 76 consecutive premature infants of <30 weeks GA or <1000 g born between 1 January 2001 and 31 August 2004 in the neonatal intensive care unit of Hôpital Saint Brieuc (France). Fifteen of these very premature infants were not included in the study: six failed to survive beyond the first months of life and the monitoring of an additional nine could not be completed by the investigators.

What is already known on this topic

  • Very preterm infants have a poor neurological prognosis despite recent advances in neonatal intensive care.

  • EEG is a useful tool to predict neuromotor disability in preterm infants.

What this study adds

  • EEG is a sensitive method to predict both neuromotor disability and cognitive outcome in very preterm infants.

  • Dysmaturity is the predominant EEG pattern found in infants who later develop severe or moderate sequelae.


EEGs were performed either with a Neurofax EEG-7310F electroencephalograph (Nihon-Kohden, Tokyo, Japan) or with DG32 Medelec or Medatec digital devices (Vickers Medical, Old Woking, Surrey, UK). Nine bipolar conduction electrodes were placed according to the modified 10–20 international system (Fp1, Fp2, T3, T4, C3, C4, O1, O2, Cz). The time constant was 0.3 s, a low pass filter was set at 70 Hz, and amplitudes were 10 μV/mm. ECG and respiration were systematically monitored, as was a video when digital equipment was used. Any use of sedative medications was carefully noted. The mean duration of the recordings was 60 min, with a minimum of 45 min in order to include the different sleep states.

In our hospital, and in keeping with previous guidelines,15 most initial EEGs were routinely recorded in 1 week after birth, usually within the first 4 days. Additional EEG recordings were performed at 31–32 and 36 weeks GA when possible or earlier depending on clinical state.

Each EEG was reviewed by two of three trained neurologists (KB, ALB, AF), working together without knowledge of the child's outcome. These EEGs were analysed according to background activity and the occurrence of other abnormal features. Analysis of background activity included: symmetry (an interhemispheric voltage difference greater than 50% was considered abnormal), spatial organisation, vigilance states, activity bursts, duration of discontinuous periods, morphology of δ and θ waves and the presence of δ brushes and temporal sawtooth waveform. Regarding the duration of discontinuous periods (ie, cerebral activity <10 µV), a maximum interburst interval (IBI) was considered abnormal if it was longer than 40 s above 28 weeks GA, 20 s above 30 weeks GA, 15 s above 32 weeks GA and 10 s above 34 weeks GA.16 The abnormal EEGs were then classified as disorganised (asymmetry or deformed morphology of δ waves and brushes), dysmature (prolonged IBI, reduction or absence of patterns expected at a given GA, or persistence in older infants of premature features), or showing other abnormal features including electrical seizures (ie, rhythmic discharges lasting at least 10 s),17 18 brief rhythmic discharges (defined as discharges less than 10 s) and non-ictal patterns as positive rolandic sharp waves (PRS) type A (more than 1/min), temporal positive slow waves (PTS), and repetitive transient sharp or slow waves (focal or not). Abnormal background activity was not considered if sedative medications were given.


In our institution, families were systematically offered clinical follow-up at the CAMSP, a centre providing health and social care for children under 6 years of age. A neuropaediatrician reviewed the children at 4 and 9 months corrected ages, and then at 18 months, 3–4 years and 5–6 years. The physician evaluated language as well as gross and fine motor skills through a standardised interview and a complete clinical examination. If a deficit was detected, specific tests were carried out for oral language and cognitive skills development (French batteries ERTL419 and BREV20) between the ages of 4 and 6.

Statistical analysis

True positives (TPs) were defined as infants with severe or moderate sequelae and at least one abnormal EEG, true negatives (TNs) were infants without sequelae and with normal EEGs, false positives (FPs) were infants without sequelae and at least one abnormal EEG, and false negatives (FN) were infants with severe or moderate sequelae and without abnormal EEG. The sensitivity was calculated as TP/(TP+FN), the specificity as TN/(FP+TN), the positive predictive value as TP/(TP+FP) and the negative predictive value as TN/(TN+FN).



The 61 infants (33 males) had a mean follow-up of 5.6 years (range 1.9–8.4 years), their mean GA was 28.5 weeks (range 25.1–30.8 weeks) and their mean birth weight was 1114 g (range 580–1580 g) (table 1).

Table 1

Duration of follow-up, number of EEGs and age at the first recording in relation to gestational age

Children were classified after clinical notes were reviewed. Clinical examinations and developmental tests showed that 25 infants had normal development. Seven infants had motor sequelae: hemiparesis (2), spastic diplegia (2), spastic tetraplegia (1), cerebral palsy with dystonia (1) and moderate distal spasticity (1). The 29 others had delayed psychomotor development involving gross motor skills (7), fine motor dexterity (26), language (18) or self-care skills (8). When these results were considered, children with abnormal development could be divided in two groups. Eleven had severe sequelae, six with global developmental delays and five with major motor handicaps which precluded attendance at an ordinary school, in contrast to 25 children who had moderate sequelae, including right hemiparesis (1), attention deficit (7), hyperactivity (6), speech disorders (1), dyspraxia (10), or speech disorders and dyspraxia (11) (table 2).

Table 2

Psychomotor outcome of children in relation to gestational age


Of a total of 187 EEGs performed in these 61 preterm newborns, 53 were abnormal, in 33 infants. Several abnormalities were frequently found in the same recording. Forty-three EEGs (23 infants) were dysmature, showing a prolonged IBI on five recordings, and the absence of EEG features expected at the GA (with a lag time of at least 2 weeks) or persistence at an older age of premature features in 41. Thirteen EEGs (12 infants) were disorganised, showing an abnormal background activity with deformed morphology of δ waves and brushes. Two EEGs displayed electrical seizures: a right central rhythmic discharge of 25 s and a right temporal rhythmic discharge of 15 s. Fifteen EEGs showed other abnormal features (8 infants): transient right hemispheric θ sharp waves (1 infant, 3 recordings), a temporal brief rhythmic discharge of 8 s (1 infant, 1 recording), transient central, frontal or temporal θ sharp waves (3 infants, 6 recordings), diffuse repetitive transient θ slow waves (2 infants, 2 recordings), occipital repetitive transient θ slow waves (1 infant, 1 recording) and abundant PRS (2 infants, 2 recordings) (table 3). Seven EEGs (5 infants) showed non-significant (<1/min) PRS and 26 EEGs (21 infants) showed PTS.

Table 3

EEG characteristics in relation to psychomotor outcome of infants

Correlations between EEGs and long-term outcome

All EEGs were normal for 22 of the 25 children with normal development. In the three others, a dysmature pattern was noted in only one EEG, but in these three recordings quiet sleep could not be obtained.

Thirty-six preterm newborns developed sequelae. In this group, 30 had at least one abnormal EEG and for 18 the first EEG was normal. Five had two abnormal EEGs, six had three abnormal EEGs and one had four abnormal EEGs. Dysmaturity was the predominant pattern found in infants with severe or moderate sequelae, in 19 and 24 EEGs, respectively. This pattern of dysmaturity was persistent on several EEGs in 12 newborns who all developed moderate or severe impairment. Of the 12 infants with a disorganised pattern, all had cognitive sequelae and two presented with cerebral palsy.

Among the six newborns with sequelae and normal recordings, two had severe sequelae: one had spastic diplegia while the other had global developmental delay. The other four infants had moderate sequelae: three had dyspraxia, one had a language disorder and two had attention disorders. The first EEG was performed after the 5th day of life in four of these six newborns and after the 13th day of life in the remaining two. In one infant IBI was 40 s at 29 weeks GA and non-significant PRS and PTS were recorded on several EEGs in three infants.

In two children with leukomalacia and severe outcome, recordings showed significant PRS, combined with a disorganised pattern. Electrical seizures without clinical manifestations were found in two neonates: one developed left hemiparesis, while the other had mild dyspraxia and hyperactivity on clinical examination at 4.7 years of age. Abnormal transients or rhythmic discharges shorter than 10 s were observed in eight infants who all developed moderate or severe sequelae. In seven of these, these transient abnormalities were associated with dysmature or disorganised patterns.

Taking these results as a whole (children with moderate and severe sequelae have been pooled together), the sensitivity of EEG to predict psychomotor outcome was 83.3%, the specificity 88%, the positive predictive value 90.9% and the negative predictive value 78.6%.


The neurological prognosis of very preterm infants remains poor despite advances in neonatal intensive care in industrialised countries.1 7 During the last 50 years, in contrast to mortality rates which have continued to decrease, the long term cerebral outcome and neurological morbidity of these infants seems to have worsened.5 7 21 In our series only 25 (41%) of the very preterm neonates had a normal developmental outcome, while 11 (18%) had severe sequelae. This is consistent with several studies on the same population regarding GA.1 21 22 More recent studies4 23 reported a higher proportion of normal outcomes but with a shorter follow-up period than in the present work (median 5.6 years), making comparison difficult since some sequelae emerge later, in particular cognitive impairments.

It is important to note that we did not include acute-stage EEG abnormalities (consisting of changes in continuity, frequency and amplitude immediately after a brain injury),6 8 but concentrated instead on chronic abnormalities because the latter appeared to be closely related to neurodevelopmental outcome,9 14 the main focus of this study. Furthermore, in the correlation analysis we did not take into account several EEG features because their clinical significance was questionable (like PTS)24,,26 or because they have been described since we started our study (like occipital and frontal sharp waves).18 27

It is generally assumed that severe periventricular-intraventricular haemorrhage (PIVH) and periventricular leukomalacia (PLM) are responsible for most neurological sequelae.9 12 PRS were first described on neonatal EEG by Cukier et al28 in 1972, and subsequently well correlated with deep white matter injury.12 29,,31 However, we found very few significant PRS (>1/min) in this series, and only three infants had leukomalacia. This discrepancy requires further longitudinal studies to ascertain if it is the result of dramatic improvements in perinatal care, with a substantial decrease in PLM.32

Similarly, electrical seizures, often subclinical, are usually associated with a poor prognosis,11 33 in particular when they arise on an abnormal background EEG.34 Seizures are frequent in neonates who develop PIVH or PLM13 and Marret et al11 observed seizures especially in infants who died with diffuse cerebral injury or primary periventricular injury. Here we found electrical seizures without clinical manifestations in only two children, one of whom developed left hemiparesis. In contrast, a favourable outcome may be expected in cases with well-preserved physiological rhythms in the background EEG,17 34 as in our second patient who had only mild dyspraxia and hyperactivity at 4.7 years of age.

Thirteen recordings (8 infants) included abnormal transients or rhythmic discharge. For some authors35 these abnormal features have no prognostic significance in preterm infants when EEGs are recorded before term age, but for others36 brief rhythmic discharges of <10 s are associated with an increased risk of abnormal neurodevelopment. In the present work these abnormal features were seen in infants with poor outcome. In one of these children, who developed moderate dyspraxia at the age of 6, an 8 s θ rhythmic discharge was the sole abnormality recorded. Nonetheless, their predictive value, taken separately, remains uncertain, as the seven other infants presented a combination of abnormal transients with either dysmaturity or disorganised EEG patterns.

Our major finding in this study is the strong correlation between background EEG abnormalities and poor neurological outcome, in keeping with former reports.6 8 9 11 14 30 34 37,,39 Several authors6 9 14 40 have described an association between disorganised patterns and deep white-matter lesions as well as between cortical involvement and dysmature patterns. In the study by Okumura et al,9 a severe disorganised pattern was associated with moderate or severe cerebral palsy in 75% of their patients, 50% of whom had mental retardation, but a significant number of children were lost during follow-up, which did not exceed 3 years. In our study we found a disorganised pattern in 12 infants, two of whom had cerebral palsy and all of whom had cognitive sequelae. The longer follow-up of our patients probably explains this high level of cognitive sequelae, which are more accurately identified at an older age.

A dysmature pattern most likely reflects a cortical injury.6 9 14 39 41 Moreover, our study confirmed that a dysmature pattern identified on several EEGs (12 newborns) was always linked to a poor outcome.6 37 However, the dysmature pattern may also be transient and appear on only one EEG. In this case, its significance is uncertain: among 11 infants with a transient pattern of dysmaturity, three had normal development. When the characteristics of these dysmature tracings were examined, prolonged IBI was observed in five of our infants with a poor outcome, which is in keeping with other series,38 41 however the absence of EEG features expected at a given GA (with a lag time of at least 2 weeks) was in this study, the predominant pattern found in the 20 children with severe or mild sequelae. These results should prompt the interpreting physician to pay close attention to the EEG features of brain maturation in preterm newborns.

Regarding psychomotor development prognosis, the sensitivity and specificity of EEG in our study are comparable to the results of previous reports,9 42 with a rather high (90.9%) positive predictive value. This indicates that an infant with an abnormal EEG is at high risk of developing moderate or severe sequelae. Several authors have highlighted the sensitivity of early postnatal recordings for the prediction of neurological outcome.6 8 Our findings confirmed these conclusions as five out of six FN patients received their first EEG very late, after 5–15 days of life. But normal tracings within the first week of life are not always correlated with a good outcome, and some features, such as abnormal transients, may appear higher at term age.22 35 In our series, 18 of the 30 TP children had a normal first EEG. Furthermore, dysmaturity features may be transient.22 So, one early EEG is often insufficient, with additional recordings being necessary to detect genuine high-risk newborns and to forecast neurological outcomes.10 13 18

In summary, this study carried out on 61 very premature newborns with a mean follow-up of 5.6 years supports the use of EEG as a sensitive method for assessing the cognitive and neuromotor prognosis in this group. Dysmaturity was the predominant pattern found in infants with severe or moderate sequelae. We emphasise the usefulness of an early postnatal tracing although in many cases abnormal patterns are transient and serial recordings are necessary to improve the prognostic value of EEG.


The authors are grateful to Marie-Dominique Lamblin for her valuable suggestions and JF Vernet for help with the English.


View Abstract


  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the Ethics Committee, CHU de Saint Etienne.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.