EEG for Predicting Early Neurodevelopment in Preterm Infants: An Observational Cohort Study
OBJECTIVE: To clarify the prognostic value of conventional EEG for the identification of preterm infants at risk for subsequent adverse neurodevelopment in the current perinatal care and medicine setting.
METHODS: We studied 780 EEG records of 333 preterm infants born <34 weeks’ gestation between 2002 and 2008. Serial EEG recordings were conducted during 3 time periods; at least once each within days 6 (first period), during days 7 to 19 (second period), and days 20 to 36 (third period). The presence and the grade of EEG background abnormalities were assessed according to an established classification system. Neurodevelopmental outcomes were assessed at a corrected age of 12 to 18 months.
RESULTS: Of the 333 infants, 33 (10%) had developmental delay and 34 (10%) had cerebral palsy. The presence of EEG abnormalities was significantly predictive of developmental delay and cerebral palsy at all 3 time periods: the first period (n = 265; odds ratio [OR], 4.5; 95% confidence interval [CI], 2.2–9.4), the second period (n = 278; OR, 7.6; 95% CI, 3.6–16), and the third period (n = 237; OR, 5.9; 95% CI, 2.8–13). The grade of EEG abnormalities correlated with the incidence of developmental delay or cerebral palsy in all periods (P < .001). After controlling for other clinical variables, including severe brain injury, EEG abnormality in the second period was an independent predictor of developmental delay (OR, 3.2; 95% CI, 1.1–9.7) and cerebral palsy (OR, 6.8; 95% CI 2.0–23).
CONCLUSIONS: EEG abnormalities within the first month of life significantly predict adverse neurodevelopment at a corrected age of 12 to 18 months in the current preterm survivor.
- ASA —
- acute stage background EEG abnormality
- CI —
- confidence interval
- CP —
- cerebral palsy
- CSA —
- chronic stage background EEG abnormality
- DD —
- developmental delay
- IVH —
- intraventricular hemorrhage
- OR —
- odds ratio
- PRS —
- positive rolandic sharp waves
- PVL —
- periventricular leukomalacia
What’s Known on This Subject:
Previous studies suggest that abnormal findings on conventional EEG during the neonatal period are associated with death or severe brain injury in preterm infants. However, large cohort studies on preterm EEG for predicting later neurodevelopmental outcome remain scarce.
What This Study Adds:
This study demonstrates precise prognostic values of conventional EEG for predicting neurodevelopmental outcome in the current perinatal care setting. Additionally, its prognostic values are independent of severe injury on neuroimaging and clinical risk factors.
Perinatal care and medicine in preterm infants has made dramatic improvements over the past several decades. However, preterm infants remain at high risk for disabilities including cognitive, motor, behavioral, emotional, and academic challenges, which occur in >50% of children born extremely preterm.1 Early recognition of at-risk infants may improve their neurodevelopmental outcomes because several neuroprotective agents, such as caffeine and erythropoietin, are available to ameliorate cerebral abnormalities in the preterm population.2,3
Previous investigations in preterm infants have shown that neonatal EEG was predictive of neurodevelopmental outcomes.4–6 However, those results were from early studies conducted at a time when the mortality rate and incidence of severe brain injury, such as intraventricular hemorrhage (IVH) and periventricular leukomalacia (PVL), was high among preterm infants. Subsequent studies tended to focus on 1 specific characteristic of EEG abnormality to predict neurologic outcomes, particularly the positive rolandic sharp waves (PRS).7–9 However, PRS are rare in current clinical practice, and a restricted assessment may underestimate EEG abnormalities compared with a full evaluation of background EEG abnormalities by using a comprehensive classification system.
The aim of the current study was to determine the predictive value of conventional EEG for early neurodevelopmental outcomes in the current perinatal care and medicine/medical setting, with a particular focus on EEG within the first month of life.
We recruited 436 preterm infants born at 33 weeks’ gestation or less admitted to the neonatal ICU in Anjo Kosei Hospital between February 2002 and April 2008. This hospital provides tertiary level care for newborns in a district with a population of ∼800 000. In total, 103 (24%) infants were excluded from the study: 4 who had congenital chromosomal or cerebral anomalies, 8 who had no EEG recordings during the study period, 40 who were lost to follow-up, 13 who died within the first month of life, 14 who died between the second month of life and 12 months of corrected age, and 24 who had no developmental tests between 12 and 18 months of corrected age. Thus, the current study included 333 infants for whom adequate neurodevelopmental outcome data were available at 12 to 18 months of corrected age. The findings of EEG in 13 infants with PVL were reported elsewhere10 and were also included in this study.
Written informed consent for EEG recording was obtained from at least 1 parent of each infant. The study was approved by the ethics committee of Anjo Kosei Hospital.
At least 1 EEG recording was scheduled for each of 3 time periods: within 72 hours of life, between days 7 and 14, and 1 month after birth, which were assigned on admission. All EEGs were performed between 1 pm and 5 pm after feeding by using a digital EEG system (Nihon Kohden, Tokyo, Japan) and recorded polygraphically along with an electrocardiogram, respiratory movement, and electrooculogram at the bedside for >40 minutes. At least 8 electrodes were placed at AF3, AF4, C3, C4, O1, O2, T3, and T4, in accordance with the international 10 to 20 system as previously reported.10 Some of the records were not obtained as scheduled, because the infant was in an unstable condition or hospital services were not offered on the scheduled day. Thus, EEGs recorded on days 1 to 6, days 7 to 19, and days 20 to 36 were regarded as the first, second, and third period, respectively.
A standardized grading system was used to evaluate acute stage (ASA) and chronic stage (CSA) background EEG abnormalities in preterm infants as previously described.11 ASA and CSA were graded as mild, moderate, and severe (Supplemental Information).10 ASA was characterized as suppressed background activity with decreased continuity, lower amplitude, and/or attenuated fast-wave background activity. CSA was characterized by a disorganized pattern (deformed δ waves with abnormal sharp waves and abnormal brushes) or as a dysmature pattern for postmenstrual age.12,13 This classification was also applied to unilateral EEG abnormalities. Additionally, the presence or absence of paroxysmal electrographic seizure activity was evaluated. When 2 or more EEG recordings were performed in a time period, the worst ASA and CSA grades were used in the study. Two of the authors independently evaluated all EEGs within 7 days after the recording blinded to the clinical information. When the 2 observers’ judgment differed (<5% of the records), consensus was reached after a discussion.
Cranial ultrasound with use of a 7.5-MHz transducer was performed in all infants every day during the first week of life and thereafter at least twice a week until discharge as routine care by neonatologists. One of the authors performed an additional cranial ultrasound to confirm the findings assessed as abnormal or equivocal. Furthermore, 284 (85%) infants underwent MRI at term equivalent age. Two authors evaluated the cranial ultrasound and MR images for the presence of grade III/IV IVH or cystic PVL. IVH grading was defined according to the classification of Papile et al.14 Cystic PVL was defined as the presence of multiple cystic lesions ≥3 mm in size in the bilateral periventricular white matter.
The infants’ neurodevelopmental outcome was assessed between 12 and 18 months of corrected age. The outcome variables were developmental delay (DD) or cerebral palsy (CP). DD was diagnosed by using the Tsumori-Inage Infant Developmental Scale or the Kyoto Scale of Psychological Development 2001. These scales are standardized and widely used developmental tests in Japan.15,16 Significant DD was defined as a developmental quotient of <70 in each developmental test. CP was diagnosed by pediatric neurologists who had access to existing clinical data for the children with the use of standard criteria including the location of the impairment or body part affected, such as hemiplegia or diplegia, and impairment of muscle tone and reflexes; severity was assessed based on gross motor function.17 Significant DD or CP was considered to be an adverse neurodevelopmental outcome in the current study.
Assuming a 25% incidence of EEG abnormalities, we estimated a total sample of n = 250 as having 80% statistical power to detect a difference of proportion of 0.15 in outcome. The association between EEG abnormality grade and the rate of outcome measures was determined by using the χ2 test. The diagnostic accuracy of EEG abnormalities was assessed by sensitivity, specificity, and positive and negative predictive values calculated from contingency tables. Odds ratios (ORs) and 95% confidence intervals (CIs) of the χ2 analyses were a measure of the strength of associations found in the univariate analysis. A forced-entry multiple logistic regression model was used to assess the association between EEG abnormalities at each time period and clinical confounders. The following variables were entered in the model as covariants: EEG abnormalities in each period, severe brain injury detected by cranial ultrasound or MRI (ie, grade III/IV IVH or cystic PVL), gestational age at birth <28 weeks, small for gestational age, male gender, multiple birth, 5-minute Apgar score of <6, inotrope use (dopamine and/or dobutamine) of ≥5 μg/kg/min, postnatal corticosteroids, and chronic lung disease defined as the use of oxygen therapy or positive end-expiratory pressure at 36 weeks’ gestation. Logistic regression was used to assess the association between EEG abnormalities and sedative use when controlling for severe brain injury and adverse outcomes. The Statistical Package for the Social Sciences version 16.0 J for Windows (SPSS Inc., Chicago, IL) was used to conduct the statistical tests. Two-sided P values of <.05 were considered to be statistically significant.
The perinatal characteristics of the 333 infants are shown in Table 1. Gestational age ranged from 22 to 33 weeks with a mean (SD) of 30.1 (2.8) weeks. Grade III/IV IVH was found in 13 infants (4%), and cystic PVL was diagnosed in 17 (5%) infants. DD was diagnosed in 33 infants (10%), and 34 (10%) had CP at 12 to 18 months of corrected age. In total, 44 infants (13%) had an adverse outcome of either DD or CP or both.
Of the 333 infants, EEG was performed in 265 (80%) during the first period, in 278 (83%) during the second period, and in 237 (71%) during the third period. The mean (SD) day of life on which the EEG was obtained was 3.1 (1.2), 12.5 (3.5), and 28.1 (4.4) in the first, second, and third period, respectively. No significant difference in characteristics was found between the entire cohort of 333 infants and those who underwent EEG recording in each time period , with the exception that infants who received an EEG in the third period had a younger gestational age at birth (29.4 [2.7] weeks, P = .001) and smaller birth weight (1210  g, P = .005). The rate of infants with grade III/IV IVH, cystic PVL, or adverse outcomes in each period was similar to that for the total cohort.
Type or Grade of EEG Abnormality and Outcome
EEG abnormality was observed in 81 (31%), 61 (22%), and 47 (20%) infants in the first, second, and third periods, respectively. ASA accounted for 52% of the EEG abnormalities in the first period, but the rate decreased in the second (15%) and third (8%) periods. The majority of CSA was a disorganized pattern; a dysmature pattern was observed in 1 infant in the second period and in 2 other infants in third period. No infant had severe ASA, whereas 6 infants had a severely disorganized CSA pattern. The EEG abnormality grade was significantly correlated with the presence of DD or CP in all periods (all P < .001; Table 2). Electrographic seizure was observed in 2 infants: 1 infant presented CP and the other presented DD and CP.
Predictive Value for Adverse Outcomes
The sensitivity, specificity, and positive and negative predictive value of all EEG abnormalities or moderate to severe abnormalities were evaluated in each period (Table 3). The sensitivity and specificity for predicting an adverse outcome was relatively constant in each period. The sensitivity and specificity of neuroimaging abnormalities (grade III/IV IVH or cystic PVL) for predicting an adverse outcome were 0.57 (95% CI, 0.42–0.70) and 0.98 (95% CI, 0.96–0.99), respectively.
ORs for Clinical Variables and EEG Abnormalities
Several clinical variables during the neonatal period were predictive of the neurodevelopmental outcome (Table 4). Small for gestational age, a 5-minute Apgar score of <6, postnatal corticosteroids, and chronic lung disease at 36 weeks of postmenstrual age were predictive of DD, and inotrope use of ≥5 μg/kg per minute was predictive of CP. The univariate analysis revealed that abnormalities detected by neuroimaging techniques (grade III/IV IVH or cystic PVL) had the highest ORs for predicting DD and CP. Furthermore, the univariate analysis showed that the OR of EEG abnormality in any time period was significantly predictive of DD and CP. The multivariate logistic regression model revealed that the presence of EEG abnormalities in the second period was a significant predictor for DD, and that EEG abnormalities occurring in the first and second periods were significant predictors for CP independent of other clinical variables (Table 5).
Effect of Sedatives on EEG Findings
Of the 333 infants, 27 (8%) with 31 EEG records received one or more sedatives within 24 hours before the EEG recording. Of those, 19 infants received a continuous dose of midazolam (0.05–0.19 mg/kg per hour), 7 infants received a total bolus dose of pentobarbital (9–17 mg/kg per day), and 2 received a continuous dose of morphine (6–22 μg/kg per hour). Of the 31 EEGs performed under sedation, 16 (52%) showed some grade of ASA. Furthermore, we found a significant relationship between sedative use and the presence of ASA after controlling for severe brain injury and adverse outcomes (P < .001). However, exclusion of the 31 EEG records from the analyses did not change the predictive values shown in Table 3 or the ORs shown in Table 4 and did not alter our conclusions.
The current study demonstrated that EEG abnormalities within the first month of life significantly predict an adverse neurodevelopmental outcome at 12 to 18 months of corrected age in preterm infants. Although several previous studies have reported similar findings, our study has several unique characteristics: accurate predictive values in current preterm survivors, a detailed evaluation of background EEG by the use of an established classification system, and a comparison with other clinical prognostic variables.
The prognostic value of preterm EEGs changes with advances in perinatal care and medicine. The largest investigations of the prognostic value of preterm EEG, all conducted in the late 1980s and 1990s, showed that EEGs had a high prognostic value, and yet the mortality rate or the incidence of severe brain injury such as grade III/IV IVH or cystic PVL was as high as 16% to 21%.18–20 In contrast, the mortality rate in the present cohort was 6% in all live birth infants, and the incidence of severe brain injury was 9% in survivors, which is consistent with current reports on preterm survivor cohorts.21 Furthermore, the presence of PRS, a well-known prognostic marker of EEG abnormalities, was rare in our study; it was observed in 2% of survivors and in 20% of survivors with severe brain injury. PRS are highly associated with the severity of brain injury,10,22 and our finding may reflect the declining trend in severe brain injury. Thus, it is reasonable that our study would show relatively lower sensitivity and positive predictive values than previous older studies.
Under these circumstances, all of our EEG records were systematically evaluated according to an established classification system, which is critical for the assessment of EEG prognostic value. The classification system used in the current study comprised 2 EEG abnormalities, ASA and CSA, which reflect chronological changes in brain damage.11 ASA, which shows suppressed neuronal activity, is often observed within several days after brain insult, whereas CSA, which shows deranged neuronal activity modulated by abnormal signaling from impaired axons in the white matter and thalamus, is typically observed after recovery from ASA. Previous studies on preterm EEGs have focused on either the specific characteristics of ASA (eg, decreased continuity, prolonged interburst intervals)23,24 or CSA (eg, distinctive sharp waves, PRS) EEG abnormalities.7–9,25 However, evaluations focusing on 1 specific component underestimate the power of the EEG. A recent study showed that neonatal EEG, assessed by using a classification system similar to ours, predicted the neurodevelopmental outcome in 61 preterm infants assessed at a mean age of 5.6 years regardless of the age at which the EEG recording was obtained.26 The disorganized CSA pattern is observed over a longer period of time than the ASA in preterm infants. Furthermore, preterm infants exhibit disorganized CSA longer than term infants with hypoxic ischemia encephalopathy, often for 2 to 3 weeks.10 This difference between preterm and term infants may be related to their distinctive pathologies: in preterm infants, the dominant lesion is in the periventricular white matter, whereas, in term infants, the dominant lesion is located in the cortex or thalamus. As a result, the predictive values in the current study were relatively constant over the recording periods, whereas the incidence of ASA differed between the first and the following periods.
Several risk factors have been associated with subsequent adverse outcomes in preterm infants, including prematurity, small for gestational age, chronic lung disease, postnatal corticosteroid use, and brain injury27,28; however, the current study, with the use of the multivariate logistic regression model, revealed that direct assessment of the brain by using EEG is a better predictor of adverse outcomes than indirect information provided by clinical risk factors. Moreover, previous studies have shown that EEG can detect milder forms of brain injury during postnatal days 5 to 14 with a higher sensitivity than cranial ultrasound.29,30 Recent reports have shown that >50% of infants with very low birth weight have milder forms of brain injury that are detected by advanced MRI or postmortem histopathology.31 Thus, EEG provides additional information independent of other clinical variables, including severe brain injury detected by cranial ultrasound, for predicting early neurodevelopmental outcomes, particularly CP.
Thus, EEG assessment provides the earliest identification of at-risk preterm infants for future intervention. Although cranial ultrasound also contributes to early detection of IVH or PVL, the hyperechogenic characteristics are commonly observed after 24 to 48 hours after the insult. In contrast, evidence suggests that abnormal findings on conventional EEG or amplitude-integrated EEG precede those on ultrasound.32,33 Our results show that EEG has high specificity and negative predicative value even within 72 hours of life, suggesting that EEG is suitable for early screening for infants at high risk needing preventive intervention.
A strength of our study was the large number of participants and the high rate of follow-up. However, the study has several limitations. First, some infants did not undergo 3 EEGs as scheduled. The infants recorded during the third period had a younger gestational age compared with the entire cohort, because some of the infants with an older gestational age were discharged earlier than the scheduled EEG, which may have affected the interpretation of our results. Second, some infants received sedatives, which could have suppressed EEG background activity.34,35 However, in current neonatal intensive care practice, sedatives or analgesics are important agents for stabilizing general condition or relieving pain in ill preterm infants. EEG recordings obtained after the influence of sedatives may provide an accurate prediction under such conditions. Third, the current study focused on the assessment of EEG within the first month of life for early prediction. The abnormal findings and their prognostic value for EEGs taken at a later stage (eg, 40 weeks’ postmenstrual age) may be different, particularly in extremely premature infants. Finally, neurodevelopmental assessment at 12 to 18 months of corrected age may underestimate cognitive or behavioral abnormalities. Follow-up assessments at school age may be more appropriate for evaluating developmental disabilities in preterm children.
EEG abnormalities within the first month of life are associated with adverse neurodevelopmental outcomes at 12 to 18 months of corrected age in preterm infants. Although severe brain injury such as grade III/IV IVH or cystic PVL detected by cranial ultrasound or MRI during the neonatal period is the most significant marker for predicting adverse outcomes, EEG provides prognostic value independent of neuroimaging findings and clinical risk factors. A disorganized pattern without PRS is common and is a better prognostic marker in conventional EEG in the current neonatal care and medicine setting.
The authors thank the staff at Anjo Kosei Hospital, including the physicians, nurses, clinical psychologists, and clinical laboratory technicians in the Division of Electrophysiology, and, most importantly, the infants and their families who participated in this study.
- Accepted June 5, 2012.
- Address correspondence to Hiroyuki Kidokoro, MD, Department of Pediatrics, Washington University in St Louis, 660 South Euclid Ave, St Louis, MO 63110. E-mail:
All authors contributed to the planning of this study; Drs Hayashi-Kurahashi, Kidokoro, Kubota, Maruyama, and Kato assessed the cranial ultrasound, MRI, and the EEG findings; Drs Hayashi-Kurahashi and Kidokoro performed the statistical analyses; Dr Hayashi-Kurahashi wrote the first draft and Drs Kidokoro, Watanabe, and Okumura revised it.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: No external funding.
- Clancy RR,
- Tharp BR,
- Enzman D
- Lacey DJ,
- Topper WH,
- Buckwald S,
- Zorn WA,
- Berger PE
- Marret S,
- Parain D,
- Jeannot E,
- Eurin D,
- Fessard C
- Kidokoro H,
- Okumura A,
- Hayakawa F,
- et al
- Tsumori M,
- Inage N
- Ikuzawa M,
- Matsushita Y,
- Nakase A
- Biagioni E,
- Bartalena L,
- Boldrini A,
- Pieri R,
- Cioni G
- Le Bihannic A,
- Beauvais K,
- Busnel A,
- de Barace C,
- Furby A
- Short EJ,
- Klein NK,
- Lewis BA,
- et al
- Morsing E,
- Asard M,
- Ley D,
- Stjernqvist K,
- Marsál K
- Copyright © 2012 by the American Academy of Pediatrics