l:'lectroenceph alography and clinic'al Neurophysiology, 82 ( 1992 ) 197- 2(12
197
t~ 1992 Elsevier Scientific Publishers Ireland, Ltd. (/013-4649/92/$(15.[)0
EEG 90233
Sequential changes in electroencephalogram continuity in very premature infants Kazuya Goto, Koichi Wakayama, Hirotomi Sonoda and Teruyuki Ogawa Department of Pediatrics, Medical College O[ Oita, Oita 87EJ-56 (Japan) (Accepted for publication: 17 September 1991 )
Summary
This study was conducted to quantify sequential changes in electroencephalogram (EEG) continttity h~r 24 h in very premature infants. For a total of 122 days, continuous 2-channel E E G recording was conducted for 28 premature infants from 26 to 33 weeks of conceptional age (CA). None of the infants showed evidence of neurological impairment during hospitalization. Normal neurological outcome was noted at a minimum 12 months of age. By classifying each 5.5 rain epoch according to E E G continuity, the n u m b e r of contiguous epochs of each series of discontinuous type (DTs) and the n u m b e r of epochs between two series of discontinous type (IDTIs) were counted at each CA. The duration of DT decreased with increasing CA. The mean duration remained at 13-16 rain after 29 weeks CA. The mean duration of each IDTI increased with CA, up to about 1 h at 33 weeks. A constant period of DTs was noted at longer intervals with increasing CA. These changes appear related to the development of sleep state organization with CA.
Key words: Premature infants: Electroencephalogram; Discontinuous activity
Recently several investigators have studied the electroencephalogram ( E E G ) in very premature infants and have shown its usefulness as an index of cerebral function and as a means for predicting later neurological development in such infants (Hughes et al. 1983; Lombroso 1985; Torres and Anderson 1985; Benda et al, 1989; Hahn et al. 1989; Tharp et al. 1989). E E G findings on premature infants can be subdivided into two distinct patterns according to the degree of continuity: "continuous activity" and "discontinuous activity," termed tracd discontinu. From these findings, E E G continuity shows striking change with maturation and is also a sensitive marker of cerebral functioning in premature infants (Hughes et al. 1983; Anderson et al. 1985; Lombroso 1985; Torres and Anderson 1985; Benda et al. 1989; Hahn et al. 1989; Tharp et al. 1989). Tharp et al. (1989) emphasize the need for serial E E G recording to predict more precisely the outcome in high-risk premature infants. They state that the first E E G record should be obtained early in the postnatal period, since there is a high risk of events which may seriously affect the cerebral functioning in infants during this period, such as intraventricular hemorrhage and cerebral ischemia.
Correspondence to: Dr. K. Goto, D e p a r t m e n t of Pediatrics, Medical College of Oita, 1-1 ldaigaoka, Hazamamachi, Oita 879-56 (Japan). "['el.: 81o0975-49-4411.
Connell et al. (1987a) and Eyre et al. (1988) used the Oxford 4-24 continuous E E G recorder to analyze the E E G in premature infants, monitored for a prolonged period with minimal interference. They quantified EEG continuity during a 24 h period by measuring the durations of continuous and discontinuous periods and found the developmental change to accompany maturation. However, they did not study sequential changes in E E G continuity over a 24 h period. In this study, continuous E E G recording was conducted to investigate the variability in continuity over a period of 24 h, and to determine whether there is an ontogenic schedule of this variation in very premature infants.
Methods
Subject selection EEGs were obtained from a prospective study conducted from December 1986 to September 1989. The subjects were premature infants admitted to our hospital at less than 34 weeks of conceptional age (CA) who fulfilled the following criteria: (1) normal neurological examination and cranial ultrasound; (2) no evidence of asphyxia (Apgar > 5 at 5 min); (3) no congenital anomalies; (4) normal neurological outcome at a minimum 12 months of age. Gestational age was determined from the date of the mother's last menstrual period and confirmed by the Ballard scoring system. Informed consent was obtained from the parents.
198
K. G O T O ET AL. 26 weeks
Method of recording A continuous 2-channel E E G record was made over a period of at least 24 h and, if possible, up to 7 days. The recording was started within the first 24 h after birth. A bipolar montage of F3-C3 and F4-C4 using silver/silver chloride cup electrodes attached by electrode cream was applied and s k i n / e l e c t r o d e impedances were kept below 10 kS2 during the recording. The electrodes were usually changed every 6 h. If any findings of contact dermatitis due to this procedure were observed, recording was terminated. E E G records were obtained using a Biophysiological Preamplifier (Nihon Kohden AB622M). An amplifier time constant of 0.3 sec, high-frequency filter setting of 30 Hz, and amplification of 20,000 were used. The E E G signal was digitized at 20 msec interval/channel by an A-D converter (Shoei AD812) and each 24 h of data was stored in a personal computer (Panafacom C280) in the nursery. The E E G was also displayed on the the oscilloscope near the incubator. The 24 h E E G record was printed on an X-Y plotter using a compressed scale of 12 c m / m i n and 5 0 0 / x V / c m .
Method of analysis For analysis of the compressed E E G , the method of Connell et al. was used (1987a)except for a different epoch length because of our computer program, in which a sample of 1024 E E G signals is obtained in a 20 msec interval. A period of 16 samples (total time = 327.7 sec, roughly 5.5 rain) was obtained and printed per line. The whole 24 h E E G record was divided into 5.5 rain epochs, which were then classified visually into the following 5 categories according to the degree of continuity, as shown in Fig. 1: (1) Continuous type: complete continuity of activity throughout the epoch. (2) Predominantly continuous type: continuous activity for at least 80% of the epoch but with at least 2 0 0 .uV L 1 minute C F4-C4
~¢~~ ` ~ ¢ Y ` ~ v ~ I ~ . ` ' ~ ` ~ ` ~ ` ~ v ~ w ~ " ~ ` ` ~ ` ~ ` ~ ' % ~ ` ~ % ~ ` . ~ . ~ , ` ~ A' ~ , ' ; ~
PC
F 3 - C 3 ~,.~'~'~,~v.~'c,,~,w'..,,,.,,e,V,,,.,~,w~.~,-t~'~'..,-'~
•, .,'-.,,.,,~f,~,,',"~,-.,------
M F4 - C 4 ,'~-~',w.,.~'--,'-.,~.-~,--v"~,.,~/~--'~,#~,',~',,.~.~,-'--~-
#,,w,-'.'~-w,~ ---,,-
PD F 4 - C4 ~
~
-
-
-
~
4
~
\
~
-
~
D F 4 - C4 ~-~-----"~"
~,'~'~,"~--~4~
~'p"r~.~'-~
Fig. 1. Five E E G types: C, continuous; PC, predominantly continuous; M, mixed; PD, predominantly discontinuous; D, discontinuous.
29 weeks
1
2
3
4
5
6hour
s
~o.,
r
2
33 weeks
P[:)I
p
3
i
LJ
7
~
~
~,
:
r~
L~i :
Fig. 2. Representative sequences of E E G types over 6 h at 26, 29 and
33 weeks CA: C, continuous; PC, predominantly continuous; M, mixed; PD, predominantly discontinuous; D, discontinuous. The solid bar indicates a DT duration. The stippled bar indicates an IDTI duration. one interval of activity attenuated to 10-20 txV for at least 5 sec. (3) Mixed type: both continuous and discontinuous types of activity present with no clear dominance of one over the other. (4) Predominantly discontinuous type: a mainly discontinuous epoch, which nevertheless contained continuous activity for at least 20% but less than 50% of the time. It included at least one but no more than two bursts longer than 1 rain in duration. (5) Discontinuous type: activity for less than 20% of the epoch, with no burst of activity exceeding 1 min. If artifacts occupied more than 20% of an epoch, that epoch was excluded from analysis. As an example, representative sequences of epochs over 6 h at 26, 29 and 33 weeks of conceptional age (CA) are shown in Fig. 2. Changes in E E G continuity in a given period were characterized by the following measurements: (1) The number of contiguous epochs of each series of the discontinuous type (DTs). (2) The number of epochs between two series of discontinuous type (IDTIs: inter-discontinuous-type intervals). (3) The total number of epochs in each of the 5 E E G types. Following these measurements for each 24 h record, changes in measurements across CA were analyzed. The means of D T and IDT1 durations were expressed in min. The number of epochs itself was only a rough measurement, meaning that these values by no means represented details. Still, they were considered better for the understanding of the tendencies in this study.
Statistical analysis In addition to standard linear regression analysis, in which significance was set at P < 0.05, a cumulative
S E Q U E N T I A L C H A N G E S IN EEG C O N T I N U I T Y IN P R E M A T U R E INFANTS
logit analysis was used to test for categorical data which showed an ordinary response (Agresti 19841. The cumulative logit can be expressed as
TABLE I
Lj(i) = o~i + X i / ~ j
CA (weeks)
where L is the cumulative logit for over jth categories versus jth and under jth categories; j ( = 1, 2 . . . . . 1 - 1, where 1 is the number of the category) is the level of response variable; i is the level of explanatory variable corresponding to CA in this report; as are constant terms; X is the explanatory variable; /3s are parameters. If /% is a negative value, an increasing trend in response variable less than jth level is shown with increasing explanatory variable value. The statistical tests concerning the cumulative logit coefficients were performed at the significance level of P < 0.05 using test statistics (parameter value divided by standard error) referred to standard normal distribution. Analysis of the cumulative logit was conducted using statistical software developed in the Department of Computer Science and Intelligent Systems in Oita University (Ochi et al. 1992).
26 27 28
199
Number of patients, 24 h EEGs and analyzable DTs and 1DTIs at each conceptional age.
29
30 32 33 31
No. of patients
No. of 24 h EEGs
No. of DTs
No. of IDTIs
2 3 5 6 6 4 10 8
7 7 13 22 10 17 29 17
176 266 389 653 189 382 519 259
173 26(I 36~ 614 2311 363 5(18 238
so40: g g
3o-
T
,~ 2o~-
T
e.,,
10-
Results 26
Gestational ages of infants in this study ranged from 26 to 33 weeks and birth weights from 676 to 1915 g. A total of 122 24 h E E G records was obtained from the 28 babies studied. The duration of recording varied from 1 to 7 days per subject. The number of analyzable epochs varied from 204 to 263 per 24 h recording. Table I shows the numbers of patients and 24 h E E G s as well as numbers of analyzable DTs and I D T I s at each CA. When the E E G of one infant extended over two CAs, the infant was considered to belong to CA.
27
28
29 C A
30 (weeks)
31
32
33
Fig. 3. Means and S.D.s of D T d u r a t i o n s at each conceptional age.
T! i
40
¸
m
I--m
Duration of each series of the discontinuous type
2O
Fig. 3 shows the means of durations of each series of DTs (DTs duration) at each CA. The mean D T duration decreased from 24.15 _+ 17.08 rain at 26 weeks to 15.71 +_ 8.07 rain at 33 weeks. Regression analysis was
O
26
27
28
29 CA
30 (weeks)
31
32
33
Fig. 4. Means and S.D.s of IDTI durations at each conceptional age.
TABLE 11 Percentages of different DT durations at each conceptional age. Cumulative logit
DT
CA (weeks)
duration (min)
26
27
28
29
30
31
32
33
/3 *
Standard error
P value **
5.5 (%) 11.0 (%) 16.5 (%) 22.0 (%) 27.5 (%) 33.0 ~ (%)
20.5 11.9 11.4 13.7 11.9 30.6
25,2 20,3 16,2 12.4 9.0 16.9
27.5 19.0 13.0 11.0 9.0 20.5
27.4 23.0 21.7 12.7 7.7 7.5
22.8 31.2 22.8 17.5 4.2 1.5
22.0 27.7 23.3 17.8 5.8 3.4
25.4 29.1 26.6 12.3 4.4 2.2
20.3 23.4 24.9 18.4 8.4 4.6
0.0284 - 0.0569 - 0.171 0.304 0.421
0.0211 0.0183 0.0201 0.0258 0.0353
0.178 0.00184 < 0.0001 < 0.0001 < 0.0001
* /3 is a parameter value. ** P value is obtained by test statistics (/3/standard error) referred to the standard normal distribution.
200
K. G O T O lET AI..
T A B L E 111
Percentages of different IDTI durations at each conceptional age. IDTI duralion (rain) 5.5 22.0 38.5 55.0 71.5 88.0 104.5
CA (weeks) 26
C u m u l a t i v e Iogit
27
28
29
30
31
32
33
/3 *
Standard
P va[ue
* *
error ~ ~ ~ ~ ~ ~ ~
16.5 ( % ) 33.0(%) 4'-,1.5( % ) 66.0(% ) 82.5 ( % ) 99.0 (c,:'~) (%)
75.1 19.1 4.6 1.2 0 0 11
69.6 27.3 3.1 () 0 0 0
76.4 211.9 2.2 0.5 0 0 0
55.11 27.5 12.4 4.2 0.5 0.2 0.2
27.0 34.8 18.7 10.4 6.5 1.7 0.9
29.7 35.0 20.4 8.8 3.0 2.2 0.9
25.2 15.7 19.9 19.5 11.0 4.1 4.6
15.1 16.8 16.4 21.8 12.7 10.1 7.1
1t.454 11.647 0.794 1t.881 0.982 0.083
I).0224 0.0275 0.0395 0.0617 0.0976 1/.148
• 0.0001 < 0.001)1
, 0.0001 0.()001 < 0.0001 < 0.(10111
* /3 is a parameter value. **
1'
value is obtained by lest statistics ( f i / s t a n d a r d error) referred to the standard normal distribution.
conducted on all analyzed data of DT duration and CA. Linear regression analysis indicated a significant negative correlation ( r = - 0 . 2 0 ; n 2=2827; P< 0.01, regression equation y = 4 6 . 3 6 - 0 . 1 8 x ) between DT duration and CA. In spite of the overall decreasing trend in DT duration, mean values remained almost constant after 29 weeks, ranging from 13.64 rain to 15.71 min. Table II shows the percentages of different DT durations at each CA. The striking change shown in this table is that the percentages at 33 rain or more decreased up to 30 weeks and thereafter became less than 5%. The cumulative logit analysis indicated that the 5.5 rain level parameter was not statistically significant. On the other hand, for DT durations longer than 11.0 rain all /3s were negative and statistically significant. These data showed that the distribution of DT durations shifted toward 11 min and 16.5 rain with increasing CA.
was conducted between 1DTI duration and CA. A significant positive correlation was noted ( r = 0.52; n 2 2752; P < 0.001, regression equation y = - 160.99 + 6.39x). In spite of the overall increasing trend in IDTI duration, mean values obtained at 26, 27 and 28 weeks were almost thc same, being 13.96 _+ 11.35,14.58 _+ 9.27 and 13.43 _+ 9.38 min, respectively. The percentages of IDTI durations at each CA. as shown in Table 111, are presented in 3 epoch wide divisions (16.5 min length). Although the mode of IDTI duration was 5.5-16.5 min except at 3t1, 31 and 33 weeks, the percentages of the larger number of divisions increased from 29 wecks. For example, the percentages of 55 min or longer IDTls were 5.1% at 29 weeks and 51.7% at 33 weeks. In the cumulative logit analysis,/3s were positive and statistically significant at each level of IDTI duration. These results showed that the distribution of 1DTI durations shifted to kruger ones with increasing CA.
Duration between two series of discontinuous type The mean of durations of IDTIs (IDTIs duration) at each CA is shown in Fig. 4. Mean IDTI duration increased from 26 weeks to 33 weeks, from 13.96 _+ 11.35 rain to 54.70_+ 33.63 rain. Regression analysis
The percentages of each o.f the 5 EEG types The percentages of the total recording time of the 5 EEG types at each CA and the results of cumulative logit analysis are shown in Table IV. All fis of cumulative logit were negative and statistically significant.
T A B L E IV
Perccntages of each of the 5 E E G t y p e s at each conceptional age. E E G type
CA (weeks) 26
Continuous Predominantly continuous Mixed Predominantly discontinuous Discontinuous
27
Cumulative Iogit 28
29
30
31
32
33
/3 *
Standard error
P value **
-0.528
0.0127
< 0.0001
(%)
1.2
1.9
1.0
5.2
7.1
9.8
21.5
24.3
(%) (%)
6.6 9.11
9.1 13.8
8.1 15.7
17.3 19.9
19.6 21.0
20.9 22.3
28.7 17.4
29.6 15.1
I).422 0.355
0.0768 0.0667
< 1).11111)l < 1L111)01
(%)
[6.6
17.2 58.0
16.11 41.6
15.6 36.7
16.1 3t).9
12.4 21/.0
11.2 19.8
0.0669
< IL0001
66.6
19.9 55.3
- 0.322
(%)
* /3 is a parameter value. * * P value is obtained by test statistics (/3/standard error) referred to the standard normal distribution.
SEOLIENT1A1.('ttANGES IN EEG CONTINUITY IN PREMATURE INFANTS These results showed a shift of E E G types to more continuous ones with increasing CA.
Discussion Continuous E E G recording is best suited for analysis of EEG continuity in premature infants (Connell et al. 1987a,b, 1988: Eyre et al. 1988). Attention in this study was directed mainly to periods of completely discontinuous activity recorded over 24 h. The durations of DT and IDTI were measured to determine patterns in EEG continuity in very premature infants. The results arc summarized as follows: up to 28 weeks CA, DTs lasting for about 20 min appeared separated by intervals of about 15 min. Thereafter, their duration remained almost constant at about 15 min each, but the intervals between them increased with CA, up to about 1 h at 33 weeks. A constant DT duration appeared at considerably long intervals at 32 and 33 weeks CA. Eyre et al. (1988) and Connell et al. (1987a) reported the duration of runs of completely discontinuous activity by continuous E E G recording. In Eyre's study, the mean duration of a run of completely discontinuous activity decreased with CA, from 21 rain at 26 weeks to 15 min at 32 weeks. However, in their study, the number of subjects was not sufficient to confirm the duration up to 33 weeks CA. Connell et al. noted that the duration remains within the range of 5-15 min from 26 to 35 weeks. In this study, 29 or 30 weeks CA were considered the dividing point, at which the mean duration of each DT and the percentage of 33 min or kruger DTs showed significant change. The reason for this discrepancy between Connell's study and ours is unclear. Attention should also be directed to differences in the duration of DTs between Connell's study and ours. Such differences may have been due to those in the definition of the epoch and the criteria for selecting subjects. As for the difference in subjects, this study included ventilated infants, whereas Connell's study did not, so as to determine the E E G of "low risk'infants. In Eyre's study, ventilated infants were not included and periods of discontinuous activity were measured without categorizing E E G continuity. These factors may possibly have led to ambiguous results. Our criteria are considered practical for analysis of EEG continuity of very premature infants without neurological impairment since ventilator support is generally conducted for them. However, there may have been some physiological or biological instability (Watanabe et al. 1982; Tharp 1986; Greisen et al. 1987) with influence on the E E G of neonates. Further study using continuous E E G recording should thus be conducted to determine to what extent systemic disorders, including respiratory illness, affect E E G continuity.
201
The parameters used in defining sleep states were not recorded simultaneously in our study and thus it was difficult to relate our results directly to the evolution of sleep states with increasing CA. However, our results appear in agreement with other data on sleep organization and sleep E E G patterns in very, premature infants. Two stages of sleep can be recognized to some extent at 28-31 weeks CA: continuous EEG accompanied by rapid eye movement (REM) sleep and discontinuous E E G without REM (Dreyfus-Brisac 1979: Lombroso 1985; Torrcs and Anderson 1985). In our study, with increasing CA, the intervals between DTs became longer and the DT duration showed less variation. These changes seem related to sleep cycle organization. The trac{ discontinu of prematurity may possibly become the tracd alternant of non-rapid eye movement (NREM) sleep in term infants. A series of DTs, therefore, would most probably correspond to some extent to a period of N R E M sleep. The findings of Vinker (1983) and Thornberg and Thiringer (1990) on normal patterns in cerebral function monitoring in term and preterm infants agree with those of the present study. They found the duration of N R E M episodes to be 20 min, remaining remarkably constant following the start of sleep cycles in infants of 30-40 weeks CA. Trends in DTs in our study were similar to their findings, although the DT duration did not completely correspond to the duration of discontinuous activity. In the case of continuous EEG recording, the use of computerized analysis to quantify EEG continuity does not appear to present problems. If EEG continuity is in fact influenced by acute stress on cerebral function as well as structural brain lesions, real-time assessment of E E G continuity may provide valuable information for the medical management of premature infants. In continuous E E G recording with computerized analysis, quantification of sequential changes in categorized E E G continuity may prove useful and may eventually be conducted for continuous monitoring in neonatal intensive care units. This study provides important information in this respect and its application to neurologically impaired infants is now needed to assess its clinical utility. This study was supported by Grants for Scientific Research (No. 63770661 and No. 02670450) from the Ministry of Education of Japan.
References Agresti, A. Logit models for an ordinal response. In: Analysis of Ordinal Categorical Data. Wiley, New York, 1984: 12(/-137. Anderson. C.M., Torres, F. and Faoro, A. The EEG of the early premature. Electroenceph. clin. Neurophysiol., 1985, 60: 95-105.
202
Benda, G.I., Engel, R.C.H. and Zhang, Y. Prolonged inactive phases during the discontinuous pattern of prematurity in the electroencephalogram of very-low-birthweight infants. Electroenceph. clin. Neurophysiol., 1989, 72: 189-197. Connell. J.A., Oozeer, R. and Dubowitz, V. Continuous 4-channel EEG monitoring: a guide to interpretation, with normal values, in preterm infants. Neuropediatrics, 1987a, 18: 138-145. Connell, J., Oozeer, R., Regev, R., De Vries, L.S., Dubowitz, L.M.S. and Dubowitz, V. Continuous four-channel EEG monitoring in the evaluation of echodense ultrasound lesions and cystic leucomalacia. Arch. Dis. Child., 1987b, 62: 1019-1024. Connell, J.. De Vries, L., Oozeer, R., Regev, R., Dubowitz, L.M.S. and Dubowitz, V. Predictive value of early continuous electroencephalogram monitoring in ventilated preterm infants with intraventricular hemorrhage. Pediatrics, 1988, 82: 337-343. Dreyfus-Brisac, C. Neonatal electroencephalograhy. In: E.M. Scarpelli and E.V. Cosmi (Eds.), Review in Perinatal Medicine, Vol. 3. Raven Press, New York, 1979: 397-472. Eyre, J.A., Nanei, S. and Wilkinson, A.R. Quantification of changes in normal neonatal EEGs with gestation from continuous five-day recordings. Dev. Med. Child Neurol., 1988, 30: 599-607. Greisen, G., Hellstr6m-Westas, L., Lou, H., Rosdn, I. and Svcnningsen, N.W. EEG depression and germinal layer hemorrhage in the newborn, Acta Pediat. Scand., 1987, 76: 519-525. Hahn, J.S., Monyer, H. and Tharp, B.R. Interburst interval measurements in the EEGs of premature infants with normal neurological outcome. Electroenceph. clin. Neurophysiol., 1989, 73: 410418.
K. GOTO ET AL. Hughes, J.R., Fino, J. and Gagnon, L. Periods of activity and quiescence in the premature EEG. Neuropediatrics, 1983, 14: 66-72. Lombroso, C.T. Neonatal polygraphy in full-term and premature infants: a review of normal and abnormal findings. J. Clin. Neurophysiol., 1985, 2: 105-155. Ochi, Y., Sugiyama, Y. and Makino. T. Evaluation of association via the generalized linear model based on multinominal distributions. Rep. Fac. Eng. Oita Univ., 1992, 22: 49-56. Tharp, B.R. Intesive video/EEG monitoring of neonates. In: R.J. Gumnit (Ed.), Advances in Neurology, Vol. 46. Raven Press, New York, 1986: 107-112. Tharp, B.R., Scher, M.S. and Clancy, R.R. Serial EEGs in normal and abnormal infants with birth weights less than 1200 grams - a prospective study with long term lk~llow-up. Neuropediatrics, 1989, 20: 64-72. Thornberg, E. and Thiringer, K. Normal pattern of the cerebral function monitor trace in term and preterm neonates. Acta Pediat. Scand., 1990, 79: 20-25. Torres, F. and Anderson, C. The normal EEG of the human newborn. J. Clin. Neurophysiol,, 1985, 2: 89-103. Vinker, D.A. Perinatal Cerebral Function Monitoring. M.D. Thesis. University of London, 1983. Watanabe, K., Hara, K., Miyazaki, S. and Hakamada, S. Neurophysiological study of newborns with hypocalcemia. Neuropediatrics, 1982, 13: 34-38.
197
t~ 1992 Elsevier Scientific Publishers Ireland, Ltd. (/013-4649/92/$(15.[)0
EEG 90233
Sequential changes in electroencephalogram continuity in very premature infants Kazuya Goto, Koichi Wakayama, Hirotomi Sonoda and Teruyuki Ogawa Department of Pediatrics, Medical College O[ Oita, Oita 87EJ-56 (Japan) (Accepted for publication: 17 September 1991 )
Summary
This study was conducted to quantify sequential changes in electroencephalogram (EEG) continttity h~r 24 h in very premature infants. For a total of 122 days, continuous 2-channel E E G recording was conducted for 28 premature infants from 26 to 33 weeks of conceptional age (CA). None of the infants showed evidence of neurological impairment during hospitalization. Normal neurological outcome was noted at a minimum 12 months of age. By classifying each 5.5 rain epoch according to E E G continuity, the n u m b e r of contiguous epochs of each series of discontinuous type (DTs) and the n u m b e r of epochs between two series of discontinous type (IDTIs) were counted at each CA. The duration of DT decreased with increasing CA. The mean duration remained at 13-16 rain after 29 weeks CA. The mean duration of each IDTI increased with CA, up to about 1 h at 33 weeks. A constant period of DTs was noted at longer intervals with increasing CA. These changes appear related to the development of sleep state organization with CA.
Key words: Premature infants: Electroencephalogram; Discontinuous activity
Recently several investigators have studied the electroencephalogram ( E E G ) in very premature infants and have shown its usefulness as an index of cerebral function and as a means for predicting later neurological development in such infants (Hughes et al. 1983; Lombroso 1985; Torres and Anderson 1985; Benda et al, 1989; Hahn et al. 1989; Tharp et al. 1989). E E G findings on premature infants can be subdivided into two distinct patterns according to the degree of continuity: "continuous activity" and "discontinuous activity," termed tracd discontinu. From these findings, E E G continuity shows striking change with maturation and is also a sensitive marker of cerebral functioning in premature infants (Hughes et al. 1983; Anderson et al. 1985; Lombroso 1985; Torres and Anderson 1985; Benda et al. 1989; Hahn et al. 1989; Tharp et al. 1989). Tharp et al. (1989) emphasize the need for serial E E G recording to predict more precisely the outcome in high-risk premature infants. They state that the first E E G record should be obtained early in the postnatal period, since there is a high risk of events which may seriously affect the cerebral functioning in infants during this period, such as intraventricular hemorrhage and cerebral ischemia.
Correspondence to: Dr. K. Goto, D e p a r t m e n t of Pediatrics, Medical College of Oita, 1-1 ldaigaoka, Hazamamachi, Oita 879-56 (Japan). "['el.: 81o0975-49-4411.
Connell et al. (1987a) and Eyre et al. (1988) used the Oxford 4-24 continuous E E G recorder to analyze the E E G in premature infants, monitored for a prolonged period with minimal interference. They quantified EEG continuity during a 24 h period by measuring the durations of continuous and discontinuous periods and found the developmental change to accompany maturation. However, they did not study sequential changes in E E G continuity over a 24 h period. In this study, continuous E E G recording was conducted to investigate the variability in continuity over a period of 24 h, and to determine whether there is an ontogenic schedule of this variation in very premature infants.
Methods
Subject selection EEGs were obtained from a prospective study conducted from December 1986 to September 1989. The subjects were premature infants admitted to our hospital at less than 34 weeks of conceptional age (CA) who fulfilled the following criteria: (1) normal neurological examination and cranial ultrasound; (2) no evidence of asphyxia (Apgar > 5 at 5 min); (3) no congenital anomalies; (4) normal neurological outcome at a minimum 12 months of age. Gestational age was determined from the date of the mother's last menstrual period and confirmed by the Ballard scoring system. Informed consent was obtained from the parents.
198
K. G O T O ET AL. 26 weeks
Method of recording A continuous 2-channel E E G record was made over a period of at least 24 h and, if possible, up to 7 days. The recording was started within the first 24 h after birth. A bipolar montage of F3-C3 and F4-C4 using silver/silver chloride cup electrodes attached by electrode cream was applied and s k i n / e l e c t r o d e impedances were kept below 10 kS2 during the recording. The electrodes were usually changed every 6 h. If any findings of contact dermatitis due to this procedure were observed, recording was terminated. E E G records were obtained using a Biophysiological Preamplifier (Nihon Kohden AB622M). An amplifier time constant of 0.3 sec, high-frequency filter setting of 30 Hz, and amplification of 20,000 were used. The E E G signal was digitized at 20 msec interval/channel by an A-D converter (Shoei AD812) and each 24 h of data was stored in a personal computer (Panafacom C280) in the nursery. The E E G was also displayed on the the oscilloscope near the incubator. The 24 h E E G record was printed on an X-Y plotter using a compressed scale of 12 c m / m i n and 5 0 0 / x V / c m .
Method of analysis For analysis of the compressed E E G , the method of Connell et al. was used (1987a)except for a different epoch length because of our computer program, in which a sample of 1024 E E G signals is obtained in a 20 msec interval. A period of 16 samples (total time = 327.7 sec, roughly 5.5 rain) was obtained and printed per line. The whole 24 h E E G record was divided into 5.5 rain epochs, which were then classified visually into the following 5 categories according to the degree of continuity, as shown in Fig. 1: (1) Continuous type: complete continuity of activity throughout the epoch. (2) Predominantly continuous type: continuous activity for at least 80% of the epoch but with at least 2 0 0 .uV L 1 minute C F4-C4
~¢~~ ` ~ ¢ Y ` ~ v ~ I ~ . ` ' ~ ` ~ ` ~ ` ~ v ~ w ~ " ~ ` ` ~ ` ~ ` ~ ' % ~ ` ~ % ~ ` . ~ . ~ , ` ~ A' ~ , ' ; ~
PC
F 3 - C 3 ~,.~'~'~,~v.~'c,,~,w'..,,,.,,e,V,,,.,~,w~.~,-t~'~'..,-'~
•, .,'-.,,.,,~f,~,,',"~,-.,------
M F4 - C 4 ,'~-~',w.,.~'--,'-.,~.-~,--v"~,.,~/~--'~,#~,',~',,.~.~,-'--~-
#,,w,-'.'~-w,~ ---,,-
PD F 4 - C4 ~
~
-
-
-
~
4
~
\
~
-
~
D F 4 - C4 ~-~-----"~"
~,'~'~,"~--~4~
~'p"r~.~'-~
Fig. 1. Five E E G types: C, continuous; PC, predominantly continuous; M, mixed; PD, predominantly discontinuous; D, discontinuous.
29 weeks
1
2
3
4
5
6hour
s
~o.,
r
2
33 weeks
P[:)I
p
3
i
LJ
7
~
~
~,
:
r~
L~i :
Fig. 2. Representative sequences of E E G types over 6 h at 26, 29 and
33 weeks CA: C, continuous; PC, predominantly continuous; M, mixed; PD, predominantly discontinuous; D, discontinuous. The solid bar indicates a DT duration. The stippled bar indicates an IDTI duration. one interval of activity attenuated to 10-20 txV for at least 5 sec. (3) Mixed type: both continuous and discontinuous types of activity present with no clear dominance of one over the other. (4) Predominantly discontinuous type: a mainly discontinuous epoch, which nevertheless contained continuous activity for at least 20% but less than 50% of the time. It included at least one but no more than two bursts longer than 1 rain in duration. (5) Discontinuous type: activity for less than 20% of the epoch, with no burst of activity exceeding 1 min. If artifacts occupied more than 20% of an epoch, that epoch was excluded from analysis. As an example, representative sequences of epochs over 6 h at 26, 29 and 33 weeks of conceptional age (CA) are shown in Fig. 2. Changes in E E G continuity in a given period were characterized by the following measurements: (1) The number of contiguous epochs of each series of the discontinuous type (DTs). (2) The number of epochs between two series of discontinuous type (IDTIs: inter-discontinuous-type intervals). (3) The total number of epochs in each of the 5 E E G types. Following these measurements for each 24 h record, changes in measurements across CA were analyzed. The means of D T and IDT1 durations were expressed in min. The number of epochs itself was only a rough measurement, meaning that these values by no means represented details. Still, they were considered better for the understanding of the tendencies in this study.
Statistical analysis In addition to standard linear regression analysis, in which significance was set at P < 0.05, a cumulative
S E Q U E N T I A L C H A N G E S IN EEG C O N T I N U I T Y IN P R E M A T U R E INFANTS
logit analysis was used to test for categorical data which showed an ordinary response (Agresti 19841. The cumulative logit can be expressed as
TABLE I
Lj(i) = o~i + X i / ~ j
CA (weeks)
where L is the cumulative logit for over jth categories versus jth and under jth categories; j ( = 1, 2 . . . . . 1 - 1, where 1 is the number of the category) is the level of response variable; i is the level of explanatory variable corresponding to CA in this report; as are constant terms; X is the explanatory variable; /3s are parameters. If /% is a negative value, an increasing trend in response variable less than jth level is shown with increasing explanatory variable value. The statistical tests concerning the cumulative logit coefficients were performed at the significance level of P < 0.05 using test statistics (parameter value divided by standard error) referred to standard normal distribution. Analysis of the cumulative logit was conducted using statistical software developed in the Department of Computer Science and Intelligent Systems in Oita University (Ochi et al. 1992).
26 27 28
199
Number of patients, 24 h EEGs and analyzable DTs and 1DTIs at each conceptional age.
29
30 32 33 31
No. of patients
No. of 24 h EEGs
No. of DTs
No. of IDTIs
2 3 5 6 6 4 10 8
7 7 13 22 10 17 29 17
176 266 389 653 189 382 519 259
173 26(I 36~ 614 2311 363 5(18 238
so40: g g
3o-
T
,~ 2o~-
T
e.,,
10-
Results 26
Gestational ages of infants in this study ranged from 26 to 33 weeks and birth weights from 676 to 1915 g. A total of 122 24 h E E G records was obtained from the 28 babies studied. The duration of recording varied from 1 to 7 days per subject. The number of analyzable epochs varied from 204 to 263 per 24 h recording. Table I shows the numbers of patients and 24 h E E G s as well as numbers of analyzable DTs and I D T I s at each CA. When the E E G of one infant extended over two CAs, the infant was considered to belong to CA.
27
28
29 C A
30 (weeks)
31
32
33
Fig. 3. Means and S.D.s of D T d u r a t i o n s at each conceptional age.
T! i
40
¸
m
I--m
Duration of each series of the discontinuous type
2O
Fig. 3 shows the means of durations of each series of DTs (DTs duration) at each CA. The mean D T duration decreased from 24.15 _+ 17.08 rain at 26 weeks to 15.71 +_ 8.07 rain at 33 weeks. Regression analysis was
O
26
27
28
29 CA
30 (weeks)
31
32
33
Fig. 4. Means and S.D.s of IDTI durations at each conceptional age.
TABLE 11 Percentages of different DT durations at each conceptional age. Cumulative logit
DT
CA (weeks)
duration (min)
26
27
28
29
30
31
32
33
/3 *
Standard error
P value **
5.5 (%) 11.0 (%) 16.5 (%) 22.0 (%) 27.5 (%) 33.0 ~ (%)
20.5 11.9 11.4 13.7 11.9 30.6
25,2 20,3 16,2 12.4 9.0 16.9
27.5 19.0 13.0 11.0 9.0 20.5
27.4 23.0 21.7 12.7 7.7 7.5
22.8 31.2 22.8 17.5 4.2 1.5
22.0 27.7 23.3 17.8 5.8 3.4
25.4 29.1 26.6 12.3 4.4 2.2
20.3 23.4 24.9 18.4 8.4 4.6
0.0284 - 0.0569 - 0.171 0.304 0.421
0.0211 0.0183 0.0201 0.0258 0.0353
0.178 0.00184 < 0.0001 < 0.0001 < 0.0001
* /3 is a parameter value. ** P value is obtained by test statistics (/3/standard error) referred to the standard normal distribution.
200
K. G O T O lET AI..
T A B L E 111
Percentages of different IDTI durations at each conceptional age. IDTI duralion (rain) 5.5 22.0 38.5 55.0 71.5 88.0 104.5
CA (weeks) 26
C u m u l a t i v e Iogit
27
28
29
30
31
32
33
/3 *
Standard
P va[ue
* *
error ~ ~ ~ ~ ~ ~ ~
16.5 ( % ) 33.0(%) 4'-,1.5( % ) 66.0(% ) 82.5 ( % ) 99.0 (c,:'~) (%)
75.1 19.1 4.6 1.2 0 0 11
69.6 27.3 3.1 () 0 0 0
76.4 211.9 2.2 0.5 0 0 0
55.11 27.5 12.4 4.2 0.5 0.2 0.2
27.0 34.8 18.7 10.4 6.5 1.7 0.9
29.7 35.0 20.4 8.8 3.0 2.2 0.9
25.2 15.7 19.9 19.5 11.0 4.1 4.6
15.1 16.8 16.4 21.8 12.7 10.1 7.1
1t.454 11.647 0.794 1t.881 0.982 0.083
I).0224 0.0275 0.0395 0.0617 0.0976 1/.148
• 0.0001 < 0.001)1
, 0.0001 0.()001 < 0.0001 < 0.(10111
* /3 is a parameter value. **
1'
value is obtained by lest statistics ( f i / s t a n d a r d error) referred to the standard normal distribution.
conducted on all analyzed data of DT duration and CA. Linear regression analysis indicated a significant negative correlation ( r = - 0 . 2 0 ; n 2=2827; P< 0.01, regression equation y = 4 6 . 3 6 - 0 . 1 8 x ) between DT duration and CA. In spite of the overall decreasing trend in DT duration, mean values remained almost constant after 29 weeks, ranging from 13.64 rain to 15.71 min. Table II shows the percentages of different DT durations at each CA. The striking change shown in this table is that the percentages at 33 rain or more decreased up to 30 weeks and thereafter became less than 5%. The cumulative logit analysis indicated that the 5.5 rain level parameter was not statistically significant. On the other hand, for DT durations longer than 11.0 rain all /3s were negative and statistically significant. These data showed that the distribution of DT durations shifted toward 11 min and 16.5 rain with increasing CA.
was conducted between 1DTI duration and CA. A significant positive correlation was noted ( r = 0.52; n 2 2752; P < 0.001, regression equation y = - 160.99 + 6.39x). In spite of the overall increasing trend in IDTI duration, mean values obtained at 26, 27 and 28 weeks were almost thc same, being 13.96 _+ 11.35,14.58 _+ 9.27 and 13.43 _+ 9.38 min, respectively. The percentages of IDTI durations at each CA. as shown in Table 111, are presented in 3 epoch wide divisions (16.5 min length). Although the mode of IDTI duration was 5.5-16.5 min except at 3t1, 31 and 33 weeks, the percentages of the larger number of divisions increased from 29 wecks. For example, the percentages of 55 min or longer IDTls were 5.1% at 29 weeks and 51.7% at 33 weeks. In the cumulative logit analysis,/3s were positive and statistically significant at each level of IDTI duration. These results showed that the distribution of 1DTI durations shifted to kruger ones with increasing CA.
Duration between two series of discontinuous type The mean of durations of IDTIs (IDTIs duration) at each CA is shown in Fig. 4. Mean IDTI duration increased from 26 weeks to 33 weeks, from 13.96 _+ 11.35 rain to 54.70_+ 33.63 rain. Regression analysis
The percentages of each o.f the 5 EEG types The percentages of the total recording time of the 5 EEG types at each CA and the results of cumulative logit analysis are shown in Table IV. All fis of cumulative logit were negative and statistically significant.
T A B L E IV
Perccntages of each of the 5 E E G t y p e s at each conceptional age. E E G type
CA (weeks) 26
Continuous Predominantly continuous Mixed Predominantly discontinuous Discontinuous
27
Cumulative Iogit 28
29
30
31
32
33
/3 *
Standard error
P value **
-0.528
0.0127
< 0.0001
(%)
1.2
1.9
1.0
5.2
7.1
9.8
21.5
24.3
(%) (%)
6.6 9.11
9.1 13.8
8.1 15.7
17.3 19.9
19.6 21.0
20.9 22.3
28.7 17.4
29.6 15.1
I).422 0.355
0.0768 0.0667
< 1).11111)l < 1L111)01
(%)
[6.6
17.2 58.0
16.11 41.6
15.6 36.7
16.1 3t).9
12.4 21/.0
11.2 19.8
0.0669
< IL0001
66.6
19.9 55.3
- 0.322
(%)
* /3 is a parameter value. * * P value is obtained by test statistics (/3/standard error) referred to the standard normal distribution.
SEOLIENT1A1.('ttANGES IN EEG CONTINUITY IN PREMATURE INFANTS These results showed a shift of E E G types to more continuous ones with increasing CA.
Discussion Continuous E E G recording is best suited for analysis of EEG continuity in premature infants (Connell et al. 1987a,b, 1988: Eyre et al. 1988). Attention in this study was directed mainly to periods of completely discontinuous activity recorded over 24 h. The durations of DT and IDTI were measured to determine patterns in EEG continuity in very premature infants. The results arc summarized as follows: up to 28 weeks CA, DTs lasting for about 20 min appeared separated by intervals of about 15 min. Thereafter, their duration remained almost constant at about 15 min each, but the intervals between them increased with CA, up to about 1 h at 33 weeks. A constant DT duration appeared at considerably long intervals at 32 and 33 weeks CA. Eyre et al. (1988) and Connell et al. (1987a) reported the duration of runs of completely discontinuous activity by continuous E E G recording. In Eyre's study, the mean duration of a run of completely discontinuous activity decreased with CA, from 21 rain at 26 weeks to 15 min at 32 weeks. However, in their study, the number of subjects was not sufficient to confirm the duration up to 33 weeks CA. Connell et al. noted that the duration remains within the range of 5-15 min from 26 to 35 weeks. In this study, 29 or 30 weeks CA were considered the dividing point, at which the mean duration of each DT and the percentage of 33 min or kruger DTs showed significant change. The reason for this discrepancy between Connell's study and ours is unclear. Attention should also be directed to differences in the duration of DTs between Connell's study and ours. Such differences may have been due to those in the definition of the epoch and the criteria for selecting subjects. As for the difference in subjects, this study included ventilated infants, whereas Connell's study did not, so as to determine the E E G of "low risk'infants. In Eyre's study, ventilated infants were not included and periods of discontinuous activity were measured without categorizing E E G continuity. These factors may possibly have led to ambiguous results. Our criteria are considered practical for analysis of EEG continuity of very premature infants without neurological impairment since ventilator support is generally conducted for them. However, there may have been some physiological or biological instability (Watanabe et al. 1982; Tharp 1986; Greisen et al. 1987) with influence on the E E G of neonates. Further study using continuous E E G recording should thus be conducted to determine to what extent systemic disorders, including respiratory illness, affect E E G continuity.
201
The parameters used in defining sleep states were not recorded simultaneously in our study and thus it was difficult to relate our results directly to the evolution of sleep states with increasing CA. However, our results appear in agreement with other data on sleep organization and sleep E E G patterns in very, premature infants. Two stages of sleep can be recognized to some extent at 28-31 weeks CA: continuous EEG accompanied by rapid eye movement (REM) sleep and discontinuous E E G without REM (Dreyfus-Brisac 1979: Lombroso 1985; Torrcs and Anderson 1985). In our study, with increasing CA, the intervals between DTs became longer and the DT duration showed less variation. These changes seem related to sleep cycle organization. The trac{ discontinu of prematurity may possibly become the tracd alternant of non-rapid eye movement (NREM) sleep in term infants. A series of DTs, therefore, would most probably correspond to some extent to a period of N R E M sleep. The findings of Vinker (1983) and Thornberg and Thiringer (1990) on normal patterns in cerebral function monitoring in term and preterm infants agree with those of the present study. They found the duration of N R E M episodes to be 20 min, remaining remarkably constant following the start of sleep cycles in infants of 30-40 weeks CA. Trends in DTs in our study were similar to their findings, although the DT duration did not completely correspond to the duration of discontinuous activity. In the case of continuous EEG recording, the use of computerized analysis to quantify EEG continuity does not appear to present problems. If EEG continuity is in fact influenced by acute stress on cerebral function as well as structural brain lesions, real-time assessment of E E G continuity may provide valuable information for the medical management of premature infants. In continuous E E G recording with computerized analysis, quantification of sequential changes in categorized E E G continuity may prove useful and may eventually be conducted for continuous monitoring in neonatal intensive care units. This study provides important information in this respect and its application to neurologically impaired infants is now needed to assess its clinical utility. This study was supported by Grants for Scientific Research (No. 63770661 and No. 02670450) from the Ministry of Education of Japan.
References Agresti, A. Logit models for an ordinal response. In: Analysis of Ordinal Categorical Data. Wiley, New York, 1984: 12(/-137. Anderson. C.M., Torres, F. and Faoro, A. The EEG of the early premature. Electroenceph. clin. Neurophysiol., 1985, 60: 95-105.
202
Benda, G.I., Engel, R.C.H. and Zhang, Y. Prolonged inactive phases during the discontinuous pattern of prematurity in the electroencephalogram of very-low-birthweight infants. Electroenceph. clin. Neurophysiol., 1989, 72: 189-197. Connell. J.A., Oozeer, R. and Dubowitz, V. Continuous 4-channel EEG monitoring: a guide to interpretation, with normal values, in preterm infants. Neuropediatrics, 1987a, 18: 138-145. Connell, J., Oozeer, R., Regev, R., De Vries, L.S., Dubowitz, L.M.S. and Dubowitz, V. Continuous four-channel EEG monitoring in the evaluation of echodense ultrasound lesions and cystic leucomalacia. Arch. Dis. Child., 1987b, 62: 1019-1024. Connell, J.. De Vries, L., Oozeer, R., Regev, R., Dubowitz, L.M.S. and Dubowitz, V. Predictive value of early continuous electroencephalogram monitoring in ventilated preterm infants with intraventricular hemorrhage. Pediatrics, 1988, 82: 337-343. Dreyfus-Brisac, C. Neonatal electroencephalograhy. In: E.M. Scarpelli and E.V. Cosmi (Eds.), Review in Perinatal Medicine, Vol. 3. Raven Press, New York, 1979: 397-472. Eyre, J.A., Nanei, S. and Wilkinson, A.R. Quantification of changes in normal neonatal EEGs with gestation from continuous five-day recordings. Dev. Med. Child Neurol., 1988, 30: 599-607. Greisen, G., Hellstr6m-Westas, L., Lou, H., Rosdn, I. and Svcnningsen, N.W. EEG depression and germinal layer hemorrhage in the newborn, Acta Pediat. Scand., 1987, 76: 519-525. Hahn, J.S., Monyer, H. and Tharp, B.R. Interburst interval measurements in the EEGs of premature infants with normal neurological outcome. Electroenceph. clin. Neurophysiol., 1989, 73: 410418.
K. GOTO ET AL. Hughes, J.R., Fino, J. and Gagnon, L. Periods of activity and quiescence in the premature EEG. Neuropediatrics, 1983, 14: 66-72. Lombroso, C.T. Neonatal polygraphy in full-term and premature infants: a review of normal and abnormal findings. J. Clin. Neurophysiol., 1985, 2: 105-155. Ochi, Y., Sugiyama, Y. and Makino. T. Evaluation of association via the generalized linear model based on multinominal distributions. Rep. Fac. Eng. Oita Univ., 1992, 22: 49-56. Tharp, B.R. Intesive video/EEG monitoring of neonates. In: R.J. Gumnit (Ed.), Advances in Neurology, Vol. 46. Raven Press, New York, 1986: 107-112. Tharp, B.R., Scher, M.S. and Clancy, R.R. Serial EEGs in normal and abnormal infants with birth weights less than 1200 grams - a prospective study with long term lk~llow-up. Neuropediatrics, 1989, 20: 64-72. Thornberg, E. and Thiringer, K. Normal pattern of the cerebral function monitor trace in term and preterm neonates. Acta Pediat. Scand., 1990, 79: 20-25. Torres, F. and Anderson, C. The normal EEG of the human newborn. J. Clin. Neurophysiol,, 1985, 2: 89-103. Vinker, D.A. Perinatal Cerebral Function Monitoring. M.D. Thesis. University of London, 1983. Watanabe, K., Hara, K., Miyazaki, S. and Hakamada, S. Neurophysiological study of newborns with hypocalcemia. Neuropediatrics, 1982, 13: 34-38.
