Functional brain maturation in neonates as measured by EEG-sleep analyses

Clinical Neurophysiology 114 (2003) 875–882 www.elsevier.com/locate/clinph

Functional brain maturation in neonates as measured by EEG-sleep analyses Mark S. Scher*, Bobby L. Jones, Doris A. Steppe, Daniel L. Cork, Howard J. Seltman, David L. Banks Developmental Neurophysiology Laboratory, Rainbow Babies and Children’s Hospital, University Hospitals of Cleveland, Case Western Reserve University, 11100 Euclid Avenue, Cleveland, OH 44106-6090, USA Accepted 15 January 2003

Abstract Objective: Seven measures of neonatal EEG-sleep behavior were evaluated using multivariate analyses to ascertain if physiologic differences exist between healthy full- and preterm cohorts. Methods: A total of 381 24-channel EEG-sleep studies were analyzed, including 125 recordings on 50 healthy fullterm and 256 recordings on 59 asymptomatic preterm infants between 28 and 70.6 weeks post-conceptional age. One EEG study for each subject was randomly assigned (109 studies) within the time window of 38 – 44 weeks post-conceptional age. A multivariate analytic procedure was applied to the data sets, by which a ‘dysmaturity index’ was assigned for each infant, based on 7 EEG-sleep measures. This index was defined in terms of the distance from the fullterm group’s centroid (i.e. Mahalanobis distance). Receiver-operating characteristic curves (ROCs) were calculated for several different combinations of 7 EEG-sleep measures to describe differences between neonatal cohorts. Results: The ROC curve corresponding to all 7 EEG-sleep measures covered the substantially largest area among the curves for the sets of variables considered, suggesting that all 7 measures of sleep behavior were required to best discriminate between cohorts. Conclusions: This methodology exemplifies how EEG-sleep analyses can be applied to the study of functional brain maturation of infants at risk for neurodevelopment problems. Significance: Changes in EEG-sleep behavior in the preterm infant may represent altered activity-dependent development of neural circuitry, resulting in remodeling of the immature brain as a reflection of adaptation to conditions of prematurity. q 2003 International Federation of Clinical Neurophysiology. Published by Elsevier Science Ireland Ltd. All rights reserved Keywords: Electroencephalographic sleep; Neonate; Preterm; Fullterm; Brain dysmaturity

1. Introduction We have previously described differences in EEG-sleep organization between pre- and fullterm cohorts at matched post-conceptional term ages (Scher et al., 1992b, 1994a,b, c). Specific preterm EEG-sleep behaviors suggest an acceleration of brain maturation as expected for the older infant when compared to the fullterm infant. Other measures suggest a delay in brain maturation as expected for a more immature neonate (Scher, 1997). In order to reconcile physiologic precocity vs. immaturity for any particular EEG-sleep measure, we chose an analytic approach, which will assess the multifactorial definition of sleep based on * Corresponding author. Tel.: þ 1-216-844-3691; fax: þ1-216-844-8966. E-mail address: [email protected] (M.S. Scher).

time-specific relationships among diverse physiologic behaviors. Changes in behaviors with increasing postconceptional ages were also assessed. Methodologies to analyze sleep could be useful for the assessment of functional brain maturation of infants at risk for developmental disabilities (Scher et al., 1996; Whitney and Thoman, 1993), as well as specific clinical syndromes, such as sudden infant death syndrome (SIDS, Glotzbach et al., 1995). Since most infants appear neurologically asymptomatic at younger ages despite their higher risk for neurodevelopmental problems, a neurophysiologic probe might be useful as both a diagnostic and prognostic tool. This study applies multivariate analyses to describe differences in EEG-sleep behaviors between healthy preand fullterm cohorts at increasing post-conceptional ages up to 6 months of age.

1388-2457/03/$30.00 q 2003 International Federation of Clinical Neurophysiology. Published by Elsevier Science Ireland Ltd. All rights reserved doi:10.1016/S1388-2457(03)00026-9

CLINPH 2002079

876

M.S. Scher et al. / Clinical Neurophysiology 114 (2003) 875–882

2. Methods 2.1. Patient selection The clinical and demographic data for these 109 neonates included 56 female and 53 male infants. Institutional review board (IRB)-approved informed consent was obtained for all study subjects. Fifty-nine preterm infants (mean birth weight 1191 g, 788 – 1670) of # 32 weeks estimated gestational age (EGA) were recruited from a neonatal population admitted to the Neonatal Intensive Care Unit (NICU) of Magee-Women’s Hospital. Selection was based on review of maternal and neonatal medical records, combined with consultation with the attending neonatologist. All infants remained clinically asymptomatic throughout the study period. None were treated for major organ system illness (i.e. hyaline membrane disease, sepsis, etc.). No infant was medically ill during the study period including encephalopathy, seizures or systemic medical illnesses. Normal cranial ultrasounds were described for all preterm subjects. Fifty appropriate-for-gestational age fullterm infants (mean birth weight: 3556 g, 2353– 4800 g) were selected from 8 well-child nurseries. Careful review of the medical records as well as physical examinations were carried out to verify the healthy status of these fullterm infants. Infants in both cohorts were followed for at least 2 years of age and were neurodevelopmentally normal either by examination or parental report. 2.2. EEG-sleep recordings Electroencephalographic/polysomnographic studies were carried out in an environmentally controlled setting in which sound, light, humidity and tactile stimulation were monitored. All infants were studied while sleeping prone or on their sides in an open bed whichever was their usual sleeping position in the nursery. Continuous recordings began after a diaper change and feeding at 0900 – 1000 h and ended between 1200 and 1300 h on the same day. The entire 24-channel 3 h recording was digitized on a Hewlett Packard workstation (Palo Alto, CA), and these 3 h were simultaneously recorded on paper using a 21-channel electrocardiogram (EEG) machine (Nihon Kohden model 4221). One hundred twenty-five recordings were performed on the fullterm cohort, while 256 studies were recorded for the preterm group, from 28 to 70.6 weeks post-conceptional age. Only one EEG study was randomly assigned (109 studies) within the time window of 38– 44 weeks postconceptional age. Digitized neurophysiologic data for each minute of sleep during the 3 h recording were compared with the contemporaneous minute of EEG-sleep, which was visually assigned to one of 6 sleep states according to conventional neonatal EEG-sleep criteria (Pope et al., 1992, i.e. two active and two quiet sleep states, as well as indeterminate

and waking states). Monthly EEG-sleep records in the preterm infant were obtained from 28 to 70 weeks. Fullterm subjects also received monthly studies between 28 and 70 weeks. A neonatal research nurse provided clinical care for each infant during the recording session. Sleep, feeding, behavior, diaper changes, medication administration and technical comments (i.e. equipment malfunctions and environment measures, etc.) were documented on our computer database. No infants were given medications during the studies. No male children were circumcised prior to the study. 2.3. EEG-sleep measures Seven specific EEG-sleep measures were selected to quantitate the degree of EEG-sleep dysmaturity between cohorts, based on previous studies which documented differences in EEG-sleep organization between healthy pre- and fullterm cohorts at matched conceptional ages (Scher et al., 1992b, 1994a,b,c). These 7 EEG-sleep measures included arousal numbers, rapid eye movement (REM) counts, percentage of quiet sleep, sleep cycle length, spectral beta EEG energies, spectral EEG correlations between left and right centrotemporal regions (i.e. T3C3/ C4T4) and respiratory ratio (i.e. a measure of respiratory regularity). These measures were chosen because each expresses precocity or delay of preterm sleep behavior when compared with fullterm infants (Scher, 1997). Accelerated measures in the preterm group that would have been expected for the older infant, included an increased percentage of quiet sleep, longer sleep cycle lengths, increased spectral EEG correlations, decreased REMS and decreased arousals. Immature measures included decreased spectral beta EEG energies and decreased cardiorespiratory regulation, which would have been expected for a child of a younger gestational age. 2.4. Analytic methodologies The ultimate goal for this multivariate analytic procedure was to combine multiple EEG-sleep measures into a single ‘dysmaturity index’, which best expresses an infant’s EEGsleep behavior relative to the average fullterm infant. Exploratory graphical methods showed that distributions of EEG-sleep measures of the fullterm infants were well represented by normal Gaussian distributions. These normal distributions of data points justified the use of the squared Mahalanobis distance from the fullterm group center based on the fullterm group’s covariance: ðx 2 m^F Þ0

X21 F

ðx 2 m^F Þ

This index, which will be called DI, is a univariate measure of disparity of an individual from the average fullterm. If the distribution of the preterm group is more spread out, shifted

M.S. Scher et al. / Clinical Neurophysiology 114 (2003) 875–882

in position or has different orientations or shapes than the distribution of the fullterm group, we can expect that for any given DI cut-off value, the fraction of the preterm cohort exceeding that value will be greater than the number of fullterm infants exceeding that value. The DI was calculated based on a narrow time window of 38 – 44 weeks conceptional age, since significant differences between the cohorts were noted during this age range, based on exploratory univariate analyses. Given that multiple observations for a single infant were obtained during this time period, the data sets were further condensed; only one of the recordings for any particular child were (randomly) chosen during this time period. We applied an appropriate computer program from a standard statistical package (Insightful Corp., Seattle, WA), to produce a robust estimate of the covariance P matrix (Rousseeuw and van Zomeren, 1990) given by ^ 21 F . For a particular physiologic behavior of a preterm infant, x, the Mahalanobis distance to the fullterm centroid is given by X ^ 2 ¼ ðx 2 m^F Þ0 ^ 21 D ^F Þ; F ðx 2 m where, m^F is the centroid vector COR

BETA

RATIO

REM

AR

QS

CYC

0.10194

0.00872

0.7778

6.11324

0.27216

34.15946

55.24324

where, COR is the spectral correlation; BETA, spectral beta range EEG power; ratio, respiratory ratio; REM, rapid eye movements; AR, arousals; QS, quietPsleep; CYC, sleep cycle. The estimated precision matrix ^ 21 F is given by: COR

BETA

RATIO

REM

AR

QS

CYC

COR

1859.945

3785.501

994.551

2 1.882

129.532

6.346

2 0.594

BETA

3785.501

193494.283

4226.205

19.839

1732.637

17.679

2 4.035 2 0.954

RATIO

994.551

4226.205

2528.198

2 9.059

46.765

10.117

REM

2 1.882

19.839

29.059

0.111

0.641

20.041

0.004

AR

129.532

1732.637

46.765

0.641

85.159

0.122

2 0.014

QS CYC

6.346

17.679

10.117

2 0.041

0.122

0.069

2 0.007

2 0.594

24.035

20.954

0.004

2 0.014

0.007

0.004

DI distributions were then calculated for all subsets of variables that included at least two EEG-sleep measures. With DI distributions for each variable subset, comparisons were made among them. For each given ‘cut-off’ value for the DI, a calculation of ‘sensitivity’ was defined as the percentage of preterm infants in the study that are at or above the cut-off value; and the ‘specificity’ was the percentage of fullterm infants that are below the cut-off (Su and Lu, 1993). These indices were then compared based on the sensitivity for a given specificity. This approach is based on receiver-operating characteristics curves (ROC, Su and Lu, 1993). ROC curves plot sensitivity as a function of one minus specificity. This analytic procedure permits group comparisons of multiple variables. For this study, the

877

variables were represented by the DI valves for different combination EEG-sleep measures. The area under the ROC curve was used to compare dysmaturity indices between neonatal cohorts calculated from different combinations of EEG-sleep measures. Better discrimination between combinations of different measures based DI values would be represented by a larger area under the curve.

3. Results Calculations of Mahalanobis distances for all preterm subjects were compared with fullterm subjects. Table 1a and b list the 7 EEG-sleep measures for subjects in each group. Distances were shorter for the fullterm than the preterm sample. In the fullterm sample, the Mahalanobis distances ranged from 3.3 to 86.53, with a medium value of 7.79; among the preterm cohort, the range was from 4.657 to 121.5 with a mean of 19.74. The 90th quantile of distances in the fullterm cohort was 25.08 while the 95th quantile falls in the 37.58; either of these could conceivably be used as a standard to decide whether a preterm infant score is highly unusual. In total, 23 preterm children showed Mahalanobis distances that exceeded the 90th fullterm quantile, while 8 exceeded the 95th fullterm quantile. The quantiles for the squared Mahalanobis distances suggest that a lower cut-off value might be appropriate; the 90th, 95th and 99.9th quantiles are 12.02, 14.07 and 24.32, respectively (the squared Mahalanobis distance has a chisquare distribution with 7 degrees of freedom). In total, 42 preterm infants could be considered extreme when using the 90th quantile as a cut-off, while only 22 had distances above the 99.9th fullterm quantile of 24.32. Data sets for each of the 7 EEG-sleep measures were tabulated for each neonatal cohort. The same calculation of Mahalanobis distances was used for all pre- and fullterm infants. Histograms of the results of full- and preterm samples are given in Fig. 1, superimposing a kernel density estimate for the distribution of distances over the histograms. Since the above analyses suggest that the squared Mahalanobis distance is potentially a good candidate for a dysmaturity index, we applied this analytic approach to ROC curves, as described in Section 2.4. Fig. 2 shows the ROC curve that compares versions of the DI formed from different subsets of the EEG-sleep measures. The first is a ‘full’ model that includes all 7 of the EEG-sleep measures. The second subset included fewer variables, spectral correlation between two EEG channels, average spectral beta energies and average number of arousals/minute in quiet sleep were included that were chosen by exploratory univariate analyses as significant during the conceptional age range of 38 –44 weeks. A third subset depicted using the same variables as in the second, except that we replaced average number of arousals/minute with percent of time in quiet sleep. As evident from the plots, the ROC curve corresponding to the ‘full’ model (i.e.

878

M.S. Scher et al. / Clinical Neurophysiology 114 (2003) 875–882

Table 1 EEG-sleep measures in the 59 healthy preterm neonates and 50 healthy fullterm neonates GA

CA

COH

Beta

Respiratory ratio

REM

Arousals

QS

Cycle

1(a) Preterm group 26 42 26 38 26 39 26 41 27 39 27 40 27 40 27 40 27 42 28 38 28 40 28 41 28 41 28 43 28 43 28 38 28 42 28 38 28 39 28 41 28 41 28 41 28 42 28 41 28 39 28 40 28 41 28 41 28 38 29 38 29 38 29 39 29 41 29 43 29 42 29 42 29 43 30 39 30 40 30 40 30 42 30 42 30 42 30 42 30 41 30 42 30 43 30 42 30 38 31 41 32 38 32 41 32 41 32 43 32 41 32 38 35 37 35 44 31 44

0.086 0.124 0.110 0.140 0.112 0.069 0.105 0.124 0.224 0.077 0.127 0.170 0.209 0.390 0.125 0.213 0.178 0.127 0.121 0.062 0.084 0.107 0.184 0.182 0.126 0.265 0.087 0.179 0.077 0.143 0.181 0.068 0.123 0.346 0.118 0.227 0.076 0.129 0.256 0.158 0.185 0.272 0.121 0.116 0.121 0.099 0.128 0.163 0.092 0.111 0.112 0.193 0.366 0.183 0.148 0.139 0.101 0.190 0.091

0.0075 0.0080 0.0067 0.0171 0.0107 0.0051 0.0043 0.0040 0.0041 0.0131 0.0054 0.0044 0.0021 0.0024 0.0049 0.0044 0.0040 0.0054 0.0071 0.0094 0.0078 0.0057 0.0063 0.0036 0.0072 0.0061 0.0050 0.0030 0.0056 0.0082 0.0067 0.0061 0.0063 0.0031 0.0059 0.0051 0.0043 0.0066 0.0037 0.0059 0.0031 0.0063 0.0044 0.0037 0.0048 0.0059 0.0047 0.0053 0.0097 0.0062 0.0098 0.0021 0.0029 0.0041 0.0086 0.0051 0.0102 0.0014 0.0031

0.752 0.793 0.777 0.760 0.846 0.737 0.747 0.824 0.735 0.783 0.823 0.793 0.764 0.743 0.747 0.798 0.745 0.753 0.802 0.790 0.781 0.694 0.884 0.743 0.790 0.736 0.745 0.696 0.838 0.822 0.811 0.756 0.755 0.702 0.801 0.775 0.736 0.775 0.763 0.708 0.821 0.664 0.728 0.740 0.810 0.778 0.772 0.778 0.787 0.761 0.797 0.757 0.730 0.763 0.775 0.759 0.854 0.847 0.778

2.4 0.6 2.0 3.0 2.0 3.1 4.9 5.3 4.4 5.0 5.0 2.9 5.1 2.2 7.3 1.5 5.0 3.6 4.1 11.9 5.3 8.3 2.9 1.8 8.2 5.6 6.2 5.8 5.2 6.4 4.4 2.8 4.1 3.2 3.3 3.1 4.4 5.6 9.3 2.3 4.7 6.1 7.5 3.0 3.8 9.9 1.4 2.3 7.4 1.9 1.3 6.0 2.3 1.8 3.4 2.8 5.4 0.3 1.8

0.056 0.158 0.017 0.104 0.120 0.039 0.031 0.037 0.116 0.138 0.089 0.141 0.250 0.000 0.113 0.074 0.042 0.174 0.169 0.100 0.059 0.122 0.133 0.217 0.095 0.014 0.020 0.150 0.217 0.000 0.000 0.175 0.101 0.035 0.216 0.000 0.120 0.065 0.097 0.127 0.077 0.016 0.179 0.027 0.063 0.070 0.127 0.042 0.132 0.304 0.100 0.017 0.116 0.067 0.069 0.145 0.215 0.151 0.192

40.0 22.6 31.9 26.5 19.7 42.1 35.7 47.1 26.5 29.0 27.3 54.7 35.0 36.4 38.8 27.7 38.8 38.3 33.0 33.0 29.8 40.9 32.8 45.1 51.9 39.0 29.1 38.0 36.6 31.5 28.5 36.8 44.6 48.3 28.3 27.9 29.1 34.4 49.7 42.8 30.0 51.2 44.1 40.6 44.2 36.3 35.4 37.2 42.2 25.3 40.9 32.8 38.3 39.5 31.0 41.9 35.7 39.9 28.6

58 20 76 91 41 52 30 49 79 45 83 94 22 68 64 44 87 72 70 52 7 72 39 52 71 56 116 99 19 140 60 57 63 66 66 72 62 74 71 30 88 46 52 29 5 92 92 67 63 36 89 71 85 75 78 76 73

M.S. Scher et al. / Clinical Neurophysiology 114 (2003) 875–882

879

Table 1 (continued) GA

CA

1(b) Fullterm group 36 41 37 41 37 37 38 38 38 38 38 40 39 39 39 39 39 39 39 39 39 43 39 39 39 39 39 39 39 39 39 39 39 39 39 39 39 40 39 44 39 45 39 40 40 40 40 40 40 40 40 40 40 40 40 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 41 42 41 41 41 42 42 42 42 42 42 42 42 42 42 42

COH

Beta

Respiratory ratio

REM

Arousals

QS

Cycle

0.079 0.108 0.076 0.102 0.140 0.078 0.055 0.096 0.093 0.154 0.189 0.089 0.092 0.145 0.099 0.109 0.060 0.067 0.092 0.110 0.090 0.085 0.153 0.159 0.101 0.099 0.107 0.079 0.141 0.079 0.093 0.080 0.121 0.090 0.056 0.071 0.075 0.084 0.066 0.132 0.140 0.123 0.101 0.120 0.182 0.049 0.179 0.079 0.079 0.085

0.0118 0.0056 0.0085 0.0061 0.0102 0.0091 0.0111 0.0058 0.0084 0.0066 0.0069 0.0080 0.0084 0.0078 0.0082 0.0122 0.0064 0.0065 0.0070 0.0042 0.0065 0.0141 0.0107 0.0065 0.0087 0.0080 0.0135 0.0071 0.0064 0.0050 0.0084 0.0069 0.0041 0.0057 0.0063 0.0125 0.0295 0.0028 0.0064 0.0082 0.0129 0.0105 0.0078 0.0082 0.0071 0.0124 0.0099 0.0178 0.0124 0.0132

0.752 0.803 0.805 0.753 0.715 0.801 0.752 0.785 0.736 0.877 0.725 0.808 0.743 0.760 0.816 0.767 0.835 0.752 0.835 0.734 0.777 0.773 0.836 0.835 0.796 0.823 0.749 0.807 0.831 0.822 0.838 0.772 0.765 0.817 0.736 0.803 0.815 0.809 0.844 0.808 0.814 0.800 0.842 0.739 0.744 0.729 0.813 0.831 0.714 0.804

6.1 2.5 3.6 0.7 1.7 2.5 11.8 5.9 5.7 8.4 5.2 2.1 10.6 4.3 9.7 5.6 3.5 2.9 14.4 3.3 4.9 2.4 3.6 3.0 3.3 13.6 0.5 2.7 17.6 8.9 10.1 3.9 4.1 7.4 8.0 5.8 9.7 9.1 12.0 12.3 4.5 9.5 12.4 8.6 2.5 10.0 10.9 13.4 3.8 3.6

0.054 0.121 0.554 0.424 0.286 0.161 0.310 0.456 0.134 0.512 0.078 0.204 0.328 0.254 0.277 0.461 0.200 0.450 0.400 0.015 0.154 0.293 0.075 0.389 0.313 0.193 0.200 0.393 0.241 0.351 0.145 0.522 0.515 0.309 0.468 0.311 0.163 0.091 0.156 0.164 0.174 0.312 0.136 0.265 0.259 0.400 0.048 0.353 0.194 0.298

41.3 32.0 29.3 36.3 45.9 32.0 47.5 31.3 45.3 23.2 48.1 26.8 33.2 33.0 25.8 40.9 38.2 36.8 27.3 36.1 34.2 35.6 22.6 29.7 38.1 32.2 27.5 23.7 40.0 42.3 36.1 27.7 36.1 33.7 44.1 33.9 23.8 37.1 34.4 30.5 25.6 52.0 42.3 37.2 25.5 32.1 23.0 29.8 40.9 28.1

87 45 34 75 66 49 53 77 60 98 99 23 67 98 75 69 29 70 46 64 52 67 74 37 42 60 66 20 35 78 32 12 59 71 39 62 30 21 57 52 47 39 51 49 31 133 30 67 54 52

GA, gestational age at birth; CA, post-conceptional age at study time; COH, spectral measure of coherence between T3-C3/C4-T4; Beta, fraction of spectral beta energy compared to the total; Respiratory ratio, spectral measure of respiratory regularity 2 ratio of harmonics; REM, number of REM per minute; Arousals, number of arousals per minute; QS, percent of quiet sleep of the total sleep time; Cycle, length of the sleep cycle in minutes.

all 7 EEG-sleep behaviors), covers the largest area under the curve and appears to best discriminate between the full- and preterm populations. For example, the same sleep behaviors that would classify 40% of the fullterm infants as extreme

(one minus specificity), would rule approximately 75% of the preterm infants as extreme. Neither of the other two variable subsets came close to this level of discrimination; in fact, the ROC curves for the two variable groups

880

M.S. Scher et al. / Clinical Neurophysiology 114 (2003) 875–882

all 7 EEG-sleep measures best discriminate between neonatal cohorts, reflecting brain function and maturation.

4. Discussion

Fig. 1. Mahalanobis distances of sampled fullterm (upper plot) and pre-term (lower plot) observations, based on the fullterm centroid.

considered here are both very close to the 458 line that indicates no discrimination at all. Thus, it appears that removing any EEG-sleep measures would lead to loss of information regarding differences in sleep between neonatal cohorts. The distribution of the Mahalanobis distances for the preterm group around the fullterm centroids suggest that

Fig. 2. ROC curve comparing versions of the dysmaturity index formed from different subsets of the explanatory variables in which sensitivity defines the percent of preterm infants at or above the cut-off value, and the specificity defines the percent of fullterm infants below the cut-off value.

Multivariate analyses were used to compare EEG-sleep behaviors for pre- and fullterm infant groups. We used the squared Mahalanobis distance analysis procedure for the fullterm centroid as a summary calculation by which ‘dysmaturity index’ would be defined. This multivariate analysis procedure distinguished between neonatal cohorts, based on this measure of physiologic dysmaturity of the brain obtained from 7 EEG-sleep measures. It would have been difficult to gain the same amount of information between cohorts regarding sleep behaviors from a reduced set of EEG-sleep measures. ROC curve analysis illustrated that important information concerning differences in sleep organization between neonatal cohorts would be lost if all 7 EEG-sleep variables were not utilized. The concept of physiologic dysmaturity has been explored by different authors for a number of high risk neonatal groups. Earlier reports define dysmaturity only as EEG pattern changes. Holmes et al. (1979) described preterm neonates with respiratory distress syndrome who expressed transient electrographic immaturity of brain function, which resolved after their medical illness. Tharp et al. (1989) and Hahn and Tharp (1990) described neonates with chronic lung disease who expressed dysmature electrographic patterns at post-conceptional term ages, but no sleep measures were included. EEG dysmaturity for the chronic lung cohort correlated with compromised neurodevelopmental outcome at 3 years of age. Scher described both dysmature EEG and sleep characteristics in a group of neonates with chronic lung disease (Scher et al., 1992a), as well as in an asymptomatic neonatal group with prenatal substance exposure (Scher et al., 1988). Immature EEG patterns as well as altered arousal numbers and sleep architecture were noted in the chronic lung disease group. Children at risk for SIDS were described as also expressing immaturity of sleep organization and arousal mechanisms. Among the EEG-sleep measures studied, included digitized spectral EEG signals (Harper et al., 1983). We have previously defined physiologic dysmaturity for an asymptomatic preterm neonatal population (Scher et al., 1997) since preterm neonates have been shown to have a higher risk for SIDS, compared to fullterm infants (Hoffman et al., 1988; Malloy and Hoffman, 1995). Other authors suggested that some children respond to stress by acceleration rather than a delay of brain function and development. This phenomenon was described in infants with intrauterine growth retardation who exhibited advanced neurological examinations during the neonatal period (Amiel-Tison and Pettigrew, 1991), as well as shorter (i.e. precocious) brainstem evoked response latencies during infancy than expected for their post-conceptional ages

M.S. Scher et al. / Clinical Neurophysiology 114 (2003) 875–882

(Pettigrew et al., 1985). Even asymptomatic neonates exhibited precocious or advanced sleep behaviors within a day of birth (Freudigman and Thoman, 1993), with predicted delayed neurodevelopment. Environmental and biologic conditions may either accelerate or delay brain maturation, depending on which specific neuronal circuitry is affected. The infant’s altered functional expression of any brain activity reflects an adaptation to stress, to maintain homeostasis for survival (Oppenheim, 1981). This general biologic adaptive process has been more recently termed activity-dependent development and underscores the complexities of remodeling regarding neuronal circuitry, through developmental processes of dendritic arborization, synaptogenesis and apoptosis (Hughes et al., 1999). Earlier adjustments in expected remodeling, however, may prove maladaptive at later developmental stages (Oppenheim, 1981). An asymptomatic fullterm cohort, for example, who exhibited accelerated sleep behaviors at birth, may be at higher risk for developmental delay during the first years of life (Freudigman and Thoman, 1993). Neurophysiological adaptation of preterm infants who express altered sleep behaviors compared to fullterm are also associated with lower developmental scores during infancy and early childhood (Scher et al., 1996; Whitney and Thoman, 1993). EEG-sleep behaviors of infants at risk for SIDS may also illustrate this process of adaptation which results in altered functional expression of brain function and maturation, termed physiologic dysmaturity. Dysmature changes in sleep organization, arousals and cardiorespiratory patterning have been described for both, infants at risk for SIDS, as well as SIDS victims (Harper, 1986; Schechtman et al., 1992). Neuroanatomical findings at postmortem examinations document altered brain circuitry based on changes in dendritic organization, synaptogenesis and myelination (Kinney et al., 1992), which represent structural correlates of dysmaturity. Findings of increased rates of apoptosis suggest an additional strategy of increased programmed cell death, which contributes to dysmaturity, and an increased risk for SIDS (Waters et al., 1999). Physiologic brain dysmaturity as expressed during sleep may predict an increased vulnerability to SIDS of these children, when later exposed to environmental or medical stresses at older ages (Filiano and Kinney, 1994). These stresses include prone sleep positions, overheating, exposure to tobacco smoke and intercurrent respiratory infections. A higher risk for SIDS for preterm infants (Hoffman et al., 1988) compared to fullterm neonates also has been reported, especially those at younger post-conceptional ages (Harper, 1986). Several shortcomings of this study are recognized. Firstly, all preterm infants were born at # 32 weeks gestational age. Our study did not address sleep behaviors of preterm infants born at successively ages closer to term. Older gestational ages may express less physiologic dysmaturity on EEG-sleep studies. Secondly, fewer studies were obtained at older ages beyond 44 weeks conceptional

881

age; persistent sleep differences might be present at 1– 6 months of age, but were not detectable based on our limited sample sizes at these ages. Thirdly, our choice of EEG-sleep measures was based on previous studies utilizing healthy preterm infants. Postnatal medical illnesses may further alter brain function and maturation, as described for infants with chronic lung disease (Filiano and Kinney, 1994). Finally, other physiologic measurements may more accurately reflect differences in state regulation between cohorts, as recently described using instantaneous heart rate and respiratory measures (Schechtman et al., 1992). The application and testing of a quantitative algorithm such as the dysmaturity index may help identify otherwise asymptomatic infants who express altered brain function and maturation. Using this quantitative method, asymptomatic healthy preterm infants in our cohort expressed alterations in EEG-sleep state behaviors, which distinguished them from a fullterm cohort at comparable postconceptional ages. Neurophysiologic studies during sleep may offer a sensitive functional probe to assess infants at risk for a range of medical conditions from developmental disabilities to SIDS. Longitudinal EEG-sleep studies, however, are needed to substantiate the presence of dysmaturity at older ages and to investigate interactions between biologic and environmental conditions which may influence later sleep behaviors.

Acknowledgements This study was supported in part by NS01110, NS26793, NS34508, NR01894, NS41118, RR00084 to Dr Scher, the Scaife Family Foundation, The Twenty-Five Club of Magee-Womens Hospital, The Cradle Roll Auxillary and the Magee-Womens Hospital Research Fund.

References Amiel-Tison C, Pettigrew AG. Adaptive changes in the developing brain during intrauterine stress. Brain Dev 1991;13:67–76. Filiano JJ, Kinney HC. A perspective on neuropathologic findings in victims of the sudden infant death syndrome: the triple-risk model. Biol Neonate 1994;65:194–7. Freudigman KA, Thoman EB. Infant sleep during the first postnatal day: an opportunity for assessment of vulnerability. Pediatrics 1993;92:373 –9. Glotzbach SF, Ariagno RL, Harper RM. Sleep and the sudden infant death syndrome. In: Ferber R, Kryger M, editors. Principles and practice of sleep medicine in the child. Philadelphia, PA: W.B. Saunders, 1995. Hahn JS, Tharp BR. The dysmature EEG pattern in infants with bronchopulmonary dysplasia and its prognostic implications. Electroenceph clin Neurophysiol 1990;76:106– 13. Harper RM, Frostig Z, Taube D, Hoppenbrowers T, Hodgman JE. Development of sleep– waking temporal sequencing in infants at risk for the sudden infant death syndrome. Exp Neurol 1983;79:821. Harper RM. State-related physiological changes and risk for sudden infant death syndrome. Aust Pediatr J 1986;22(Suppl):55–65. Hoffman HJ, Damus K, Hillman L, Krongrad E. Risk factors for SIDS. Results of the National Institute of Child Health and Human

882

M.S. Scher et al. / Clinical Neurophysiology 114 (2003) 875–882

Development SIDS Cooperative Epidemiological Study. Schwartz PJ, Southall DP, Valdes-Dapena M, editors. Ann N Y Acad Sci 1988;533: 13–30. Holmes DL, Logan WJ, Kirkpatrick BV, Meyer EC. Central nervous system maturation in the stressed premature. Ann Neurol 1979;6: 518– 22. Hughes PE, Alexi T, Walton M, Williams CE, Drugunaw M, Clark RG, Gluckman PD. Activity and injury-dependent expression of inducible transcription factors, growth factors and apoptosis-related genes within the central nervous system. Prog Neurobiol 1999;57:421–50. Kinney H, Filiano JJ, Harper RM. The neuropathology of the sudden death syndrome. A review. J Neuropathol Exp Neurol 1992;51:155. Malloy MH, Hoffman JJ. Prematurity, sudden infant death syndrome, and age of death. Pediatrics 1995;96:464–71. Oppenheim RW. Ontogenetic adaptations and retrogressive processes in the development of the nervous system and behavior. In: Connolly K, Prechtl H, editors. Maturation and behavior development. London: William Heinnemann Publishers Spastic Society Publications, 1981. p. 73–109. Pettigrew AG, Edwards DA, Henderson-Smart DJ. The influence of intrauterine growth retardation on brainstem development of preterm infants. Dev Med Child Neurol 1985;27:467–72. Pope JE, Werner SS, Birkford RG. Atlas of neonatal electroencephalography, 2nd ed. New York, NY: Raven Press, 1992. Rousseeuw PJ, van Zomeren BC. Unmasking multivariate outliers and leverage points (with discussion). J Am Stat Assoc 1990;85:633–51. Schechtman VL, Raetz SL, Harper RK, Garfiukiel A, Wilson AJ, Southhall DP, Harper RM. Dynamic analysis of cardiac R-R intervals in normal infants and infants who subsequently succumbed to the sudden infant death syndrome. Pediatr Res 1992;31:606. Scher MS. Neurophysiological assessment of brain function and maturation. II A measure of brain dysmaturity in healthy preterm neonates. Pediatr Neurol 1997;16:287 –95. Scher MS, Richardson GA, Coble PA, Day NL, Stoffer DS. The effects of

prenatal alcohol and marijuana exposure: disturbances in neonatal sleep cycling and arousal. Pediatr Res 1988;24:101 –5. Scher MS, Richardson GA, Salerno DG, Day NL, Guthrie RD. Sleep architecture and continuity measures in neonates with chronic lung disease. Sleep 1992a;15:195 –201. Scher MS, Steppe DA, Dahl RE, Asthana S, Guthrie RD. Comparison of EEG-sleep measures in healthy full-term and preterm infants at matched conceptional ages. Sleep 1992b;15:442– 8. Scher MS, Sun M, Steppe DA, Guthrie RD, Sclabassi RJ. Comparisons of EEG spectral and correlation measures between healthy term and preterm infants. Pediatr Neurol 1994a;10:104–8. Scher MS, Sun M, Steppe DA, Banks DL, Guthrie RD, Sclabassi RJ. Comparisons of EEG state-specific spectral values between healthy full-term and preterm infants at comparable postconceptional ages. Sleep 1994b;17:47 –51. Scher MS, Steppe DA, Banks DL. Prediction of lower developmental performances of healthy neonates by neonatal EEG-sleep measures. Pediatr Neurol 1996;14:137–44. Scher MS, Steppe DA, Dokianakis SG, Sun M, Guthrie RD, Sclabassi RJ. Cardiorespiratory behavior during sleep in fullterm and preterm neonates at comparable postconceptional term ages. Pediatr Res 1994cc;36:738 –44. Su JQ, Lu JS. Linear combinations of multiple diagnostic markers. J Am Stat Assoc 1993;88:1350–5. Tharp BR, Scher MS, Clancy RR. Serial EEGs in normal and abnormal infants with birth weights less than 1200 grams – a prospective study with long term follow-up. Neuropediatrics 1989;20:815–9. Waters KA, Meehan B, Huang JQ, Gravel RA, Michaud J, Cole A. Neuronal apoptosis in sudden infant death syndrome. Pediatr Res 1999; 45:166–72. Whitney MP, Thoman EB. Early sleep patterns of premature infants and differentially related to later developmental disabilities. J Dev Behav Pediatr 1993;14:71–80.