Sleep, 8(3):193-206
© 1985 Raven Press, New York
Sleep-Wake State Organization, Neonatal
Assessment and Development in Premature
Infants During the First Year of Life. II.
Thomas F. Anders, *Marcia A. Keener, and tHelena Kraemer
Division of Child and Adolescent Psychiatry, Bradley Hospital and Brown University, Providence,
Rhode Island; *Division of Child Psychiatry and Child Development, tDepartment of Psychiatry
and Behavioral Sciences, Stanford University School of Medicine,
Stanford, California, USA
Summary: Twenty four premature infants were evaluated in their homes at seven
ages during the first year of life to determine whether sleep-wake state organization
was related to either neonatal assessment or short-term developmental outcome
measures. A model assessing environmental and biological influences on the maturational course of selected sleep-wake state parameters was also evaluated. Sleepwake state variables and neonatal assessment items were related to each other,
and both predicted developmental quotients at 6 months and 1 year of age. Concordance was present primarily in the domain of motor activities. Waking motor
behaviors and motor activity in sleep seem to independently reflect an infant's
level of developmental organization. Individual sleep-wake state variables were
influenced by both biological and environmental factors during maturation. The
developmental course of quiet sleep is primarily biologically determined, as evidenced by its relationship to the infant's birth status (gestational age or birth
weight); the infant's behavior that results in being taken from the crib during the
night, and the course of his/her sleep that occurs between midnight and morning
are dependent on both biological factors (perhaps infant irritability), and postbirth experiences (perhaps the caregiver'S response to infant irritability). The
course of active sleep and wakefulness are dependent solely on environmental
influences, and not on maturity at birth. Key Words: Sleep-wake state organization-Gestational age-Birth weight-Developmental assessment.
Numerous studies have demonstrated that infants born prematurely are at risk for disabling
developmental outcomes (1-3). Yet the process by which risk status becomes transformed
into maladaptive outcome is complex and not completely understood. Biological and environmental influences interact, some protecting the vulnerable infant and others burdening
the adaptive process and resulting in continuities and discontinuities of development (4).
The task of predicting which infant is at risk continues to be difficult (5). Pr~gnancy and
perinatal complications, neonatal behavioral and neurological evaluations, and social interactions of parent-infant dyads predict individual outcomes poorly in infants, who have
not suffered a major physiologic insult (6).
Accepted for publication June 1985.
,
Address correspondence and reprint requests to Dr. 1. F. Anders at Bradley Hospital, lOll Veterans Memorial
Parkway, East Providence, Rhode Island 02915, U.S.A.
.
193
194
T. F. ANDERS ET AL.
A necessary first step is to understand the process whereby certain behaviors or domains
of activity become organized into functional units that demonstrate continuity and predictibility. Then the stability of these units can be studied in relation to more complex levels
of organization. In early development, motor organization lends itself to study from multiple
perspectives. Animal studies have demonstrated that several discrete pontine centers are
involved in motor regulation during sleep and wakefulness (7); intrauterine fetal studies in
animals and humans have suggested that cyclic organization of body movements are present
(8,9); and motor activity in neonates has been shown to predict motor organization in
childhood (10). Thus, the maturation of the motor system merits careful scrutiny.
Since organization of sleep-wake patterns is related to central nervous system maturation,
and involves the regulation of motor activity and the inhibition of motor tone in active sleep
(AS) and quiet sleep (QS), we hypothesized that a longitudinal study of sleep-wake state
deVelopment, during the first year of life, might add to an understanding of the process
whereby risk factors become stabilized or transformed. A better understanding of this
process should improve the ability to predict and intervene effectively.
This article describes motor and mental development in a group of 24 premature infants.
The preceding article (11) described the maturation of sleep-wake state variables in this
group as compared with a group of age-matched, full-term infants. We noted that sleepwake state organization in both groups progressed comparably, suggesting that the premature
infants were not grossly abnormal.
Here we examined whether differences in developmental outcome at 6 months and 1
year of age in the premature group could be predicted from sleep-wake state organization
and/or other measures of early assessment. Specifically, we examined the relationship of
sleep-wake state organization to an obstetrical-perinatal adversity score derived from the
medical record, to neonatal assessment scores using Brazelton's neonatal behavioral assessment scale (12), and to developmental outcomes at 6 and 12 months of corrected age,
assessed by the Bayley Scales of Infant Development (13). Although the sample of preterm
infants had been especially recruited because of obstetrical and perinatal adversity factors
that presumed to place them at risk of poor outcome, it is important to note that by the
time of hospital discharge and study only a few of the subjects appeared clinically impaired.
METHODS
Subjects
The method of recruiting subjects, the socio-demographic features of the families, and
the birth status of the premature infants are described in detail in the previous article (11).
In summary, the first twenty-one families recruited agreed to participate in the study.
Initially, we had intended to study only 20 infants and had recruited one additional family
in case one family failed to finish the study. However, all of the families continued in
the study to its end. Three sets of twins were included; therefore, the premature cohort
consisted of 24 infants. The mean gestational age of the group was 31.2 weeks (SD 1.9,
range 27-35 weeks). The mean birth weight was 1,543 gm (SD, 494 gm, range 790--2,860
gm). One infant was considered small for gestational age. Birth weight and gestational age
were significantly correlated (rs 0.71, p ::;; 0.0001).
Infants entered the project at 42 weeks conceptional age (the number of weeks since
conception) and were studied on seven occasions up to 12 months corrected age (the number
of weeks from conception minus 40 weeks) in a short-term longitudinal design. All sleep
recordings were carried out at home with all-night, time-lapse video somnography. As three
Sleep, Vol. 8, No.3, 1985
SLEEP-WAKE STATES IN PREMATURE INFANTS
195
infants were still in the hospital at 42 weeks conceptional age, only 21 infants were recorded
on the initial occasion.
Assessment procedures
Pregnancy-perinatal risk. Pregnancy and perinatal histories were obtained from the
mother and from a review of hospital charts. An obstetrical-perinatal adversity score was
derived for each infant. Birth weight, gestational age, number of days in the intensive
neonatal nursery, severity of the respiratory distress syndrome, severity of neonatal jaundice
and length of time intubated were coded as mild (0 = lowest quartile), moderate (1 = middle
quartiles), or severe (2 = upper quartile); and the sum of the scores (range = 3-18)
comprised the adversity score. High- and low-risk groups of infants were formed on the
basis of their adversity scores. There were 16 infants that were considered to be at lowrisk with total scores of ~ 6 and 8 infants at high-risk with scores of ;:,: 7.
Neonatal assessment. Assessment of neonatal status was obtained by administration in
the home of Brazelton's Neonatal Behavioral Assessment Scale at 2 and 4 weeks corrected
age by one of two certified examiners (12). Examinations were scheduled so that examiners
alternated evaluations. For the most part, no infants were examined twice by the same
examiner. The Brazelton examination resembles a standardized neonatal neurological evaluation and assesses the maturity of such developmental constructs as motor organization,
state regulation, affect, perception, responsiveness, and the status of neonatal reflexes. It
is widely used as a method of testing neonatal function and the infant's interactive and
regulatory capacities. The examiners who performed the developmental assessments were
blind to the sleep-wake state organization and obstetrical-perinatal adversity scores of the
infants.
The infant's performance was summarized using two methods: one for scoring a priori
dimensions (14) and one using factor analysis (15). The former method produces four
dimensions and a summary optimality score for each infant. The dimensions are interactive
processes, motoric processes, state control, and response to stress. Factor analysis produces
seven clusters: habituation, orientation, motor behavior, range of state, regulation of state,
autonomic stability and abnormal reflexes, and a summary score. Both dimensional and
factor analysis reduce Brazelton's 27 items, scored on a scale of 1-9, to fewer variables of
reduced spread.
Sleep-wake recordings. All-night video somnograms, using time-lapse video recording
in the homes, were obtained at 2 (conceptional age 42 weeks), 4, 8, 20, 24, 36, and 52
weeks, as described in detail in the previous report. Video tapes were scored by the standard
methods also described previously (11,16,17,18). A total of 165 nights of sleep were analyzed.
Six-month and I -year assessments. Developmental outcome was measured by the Bayley
Scales of Infant Development (13). Each infant was tested on two occasions, at 24 and 52
weeks of corrected age, by one of two trained examiners. As with the Brazelton assessments,
examiners alternated evaluations so that the same infant was not examined twice by the
same person. The Bayley tests were scored in the standard manner, providing two summary
scores at each corrected age, a psychomotor developmental index, and a mental age developmental index.
Data analysis
Stability of the Brazelton assessments from 2 to 4 weeks of age, for both the dimensions
and the factors, and of the Bayley developmental quotients from 24 to 52 weeks were
Sleep, Vol. 8, No.3, 1985
196
T. F ANDERS ET AL.
determined by Spearman rank order correlation coefficients (21). Similarly, relationships
between the a priori dimensions and the factors at each age, and the relatiunships between
the Bayley mental and psychomotor quotients at each age were examined.
Because the sleep-wake state variables of any single night are unlikely to predict developmental outcome at 6 months or I year of age (22), we chose to examine developmental
profiles for each of the eight sleep-wake state variables defined in our previous report: the
percentage of active sleep (AS%), the percentage of quiet sleep (QS%), the percentage
awake (AW%) , the percentage of out-of-crib time (OOC%), the longest sustained sleep
period uninterrupted by wakefulness (LSP), the proportion of the LSP that occurs between
12 midnight and 5:00 a.m. (%LSP(12-5 a.m.», and two variables derived from a semiMarkov model, the holding time index, and the transition probability index (11,19,20). The
use of profiles emphasizes the dynamic, changing nature of sleep-wake state organization
during early development.
Since the age in weeks at the time of birth (gestational age) differs for premature infants,
for each infant, developmental profiles for each sleep-wake state variable can be constructed
using a mathematical model and statistical method developed by Kraemer et al. (23) that
is similar to multiple logistic regression analysis (:24). The method computes a regression
equation for each variable that results in three descriptors for each: the intercept reflects
the variable's performance at birth (birth status), the slope reflects the variable's rate of
development as a function of chronological age, and the root mean square error term reflects
the variable's stability around the subject's regression line. An example of a developmental
profile for LSP for an infant with the three summary descriptors is depicted in Fig. 1.
Kraemer et al. (23) speculate that maturity at birth may differentially influence the rate
of development of particular variables at subsequent corrected ages. Infants born at younger
gestational ages may, at a given corrected age, perform quite differently in some tests from
infants born at older gestational ages tested at the same corrected age. In such cases, the
younger infant's performance has not caught up. In contrast, for other variables, infants
born at younger and older gestational ages may demonstrate no differences when assessed
at a later comparable corrected age.
The model separately tests the strength of correlations between slopes and intercepts
and gestational age and birth weight, the latter two indicating the infant's biological status
at birth. For variables whose courses of development are primarily set by intrinsic central
nervous system mechanisms, and are relatively unaffected by environmental input, only
intercept coefficients will correlate significantly with gestational age and/or birth weight.
The slopes of the regression lines, for such variables, will parallel each other and not
demonstrate a significant correlation with gestational age and/or birth weight. In contrast,
for variables in which chronological age (environmental experience) significantly interacts
with biological determinants to influence the course of development, both slope and intercept
coefficients will be significantly correlated with gestational age and/or birth weight. The
slopes of the regression lines (rate of development) will then differ for each infant, demonstrating a catch-up effect. Variables whose developmental profiles are not characterized
by either intercept correlates or intercept and slope correlates of gestational age and/or
birth weight can be considered to be influenced primarily by environmental factors (23).
Application of the model provides an opportunity to examine two questions. The first
question addresses whether individual sleep-wake state variables develop differentially, some
reflecting primarily biological central nervous system programming, others more sensitive
to environmental influences. The second question addresses whether developmental profiles
Sleep, Vol. 8, No.3, 1985
197
SLEEP-WAKE STATES IN PREMATURE INFANTS
600
500
400
LSP
300
(mins)
Performance @ Birth
intercept 69.5 mins
=
200
Developmental Rate
slope = 7.94 miniwk
100
Stability
root MSE
= 122.79
OL----------------------------------------------------60
40
50
10
20
30
t
Birth @30wksGA
ehron. age - wks
x = Predicted value
•
= Actual score
FIG. 1. Developmental profile for infant BB of the sleep variable LSP [the longest single sustained period
of sleep (in minutes) during a night's recording] is portrayed. The infant's actual and predicted values derived
from the regression equation are plotted. The intercept, slope, and root mean square error (rmse) terms, derived
from the equation, are also listed. Using the developmental model of Kraemer et al. (23), intercept and slope
coefficients for all variables are correlated with birth weight and gestational age. For LSP, the infants as a group
demonstrated a weakly positive correlation with birth weight (0.35; p ~ 0.1), suggesting a biologically determined developmental course for this variable. Note also that birth at 30 weeks gestational age results in a
chronological age of 62 weeks at 52 weeks of conceptional age.
of sleep-wake state variables (the intercept, slope, and root mean square error coefficients)
predict developmental quotients at 24 and 52 weeks of age.
The Brazelton dimensions and factors and sleep-wake state regression coefficients were
entered into a stepwise mUltiple regression analysis (25) to predict Bayley mental and
psychomotor quotients at 24 and 52 weeks of age. To examine the patterns of interaction
between predictor variables and other related variables, Spearman rank order correlations
were computed (21).
The use of multiple regression analysis in this way is hazardous. The number of infants
is small and the number of variables great. Therefore, results from the multiple regression
model must be interpreted with caution. The potential importance of motor organization
as an early stable marker of development justifies its use for exploratory purposes, however.
This study, to the best of our knowledge, is the first to examine motor organization longitudinally, using sleep-wake and neurodevelopmental assessment methods. With only 24
infants it must be considered preliminary, even though 165 nights of sleep were recorded.
To be more certain about specific neonatal assessment items and sleep-wake state variables that significantly predict developmental outcome, an independent, prospective, preSleep, Vol. 8, No.3, 1985
T. F. ANDERS ET AL.
198
dictive replication study will be required. The focus of future investigations can be narrowed,
however, as a result of this pilot study.
RESULTS
Stability of neonatal behavioral assessment
Relationships between the two methods of scoring the Brazelton examination are summarized in Table 1. In examining the stability of items from 2 to 4 weeks of age, one
factor, orientation, was stable and another, motor behavior, demonstrated a trend toward
stability from 2 to 4 weeks of age (rs 0.42, p = 0.07). Two of the dimensions, motoric
processes and response to stress, were stable while two others, interactive processes and
the summary dimension, demonstrated trends toward stability (rs 0.41, p = 0.07 and rs
0.42, p = 0.06, respectively).
As expected, there was close concordance between some dimensions and factors since
many of the same items are used in both scoring methods. At both 2 and 4 weeks of age,
the orientation factor and the dimension of interactive processes were highly correlated.
The motor behavior factor was significantly correlated with the motor processes dimension
at 4 weeks only, but a trend was present at 2 weeks as well (rs - 0.42, P = 0.08). The
autonomic stability factor correlated significantly at both ages with the dimension of response
TABLE lAo Brazelton's Neonatal Behavioral
Assessment Scale (NBAS) stability from 2 to 4
weeks of age
2 to 4 weeks of age
(Spearman correlations, r,)
Factors
Orientation
Motor behavior
Dimensions
Motoric processes
Response to stress
Interactive processes
0.48"
0.42
0.50"
0.52"
0.41
TABLE lB. Comparison of dimensions and factors at
two ages
(Spearman
correlations, r,)
Factors/Dimensions
2 weeks
4 weeks
Orientation/Interactive processes
Motor behavior/Motoric processes
Autonomic stability/Response to stress
Summary
- 0.90c
-0.42
- 0.62b
-0.67 c
-0.87 c
-0.S4b
-0.67 c
-0.70c
Low scores signify optimum function for dimensions; high scores
signify optimum function for factors.
"p .:; 0.05.
b p .:; 0.01.
cp .:; 0.001.
Sleep, Vol. 8, No.3, 1985
SLEEP-WAKE STATES IN PREMATURE INFANTS
199
to stress. Finally, at each age, the summary factor was highly correlated with the summary
dimension.
We examined the two methods of scoring in yet another way. The three infants who
consistently scored lowest and the three infants who scored highest at both 24 and 52 weeks
of age on the Bayley were identified (the 6 outliers). The average (2 and 4 week) summary
dimension predicted two of the three infants with the lowest Bayley outcomes, and none
of the three infants with the highest Bayley outcomes. In contrast, the average summary
factor identified a total of four infants, including all of the infants in the high Bayley group
and one of the infants in the low Bayley group. It appears that dimensional scoring may
be more effective in identifying those infants at greater risk, whereas factor analysis may
be more effective in monitoring optimal functioning.
Maturation of sleep-wake variables
Next we examined the developmental profiles of the eight sleep-wake state variables
using the model described by Kraemer et al. (23). We correlated slope and intercept values
with gestational age and birth weight, indicators of the infant's birth status. The correlations
are presented in Table 2. The analysis suggests that at least one sleep variable, QS%,
demonstrates a course of development that is principally dependent on biological influences
(performance at birth), and does not catch up as a result of environmental influences. The
maturation of two others, the transition probability index and the LSp, also appear to reflect
biological influences (p ~ 0.1). In contrast, the developmental profiles of OOC% and
%LSP(12-5 a.m.) show patterns of development that suggest an interaction of birth status
with chronological age (p ~ 0.1). The remaining variables, AS%, AW%, and the holding
time index, seem to develop predominantly in response to environmental influences.
Bayley scales
As summarized in Table 3, the mean Bayley psychomotor quotient at 24 and 52 weeks
was 98.4 and 93, respectively. At 24 and 52 weeks of age the mean mental quotient was
104.8 and 99.3, respectively. A slight decline in developmental quotient from 6 months to
12 months is not uncommon and suggests that the Bayley test is better able to assess function
discriminatively at older ages, i.e., at older ages infants have more varied response capacities
and the Bayley examination has more varied items and hence performance can be assessed
more precisely. At 24 weeks, Bayley mental and psychomotor quotients were significantly
correlated (rs 0.74); by 52 weeks the correlation, though still significant, had diminished
(r, 0.55). These relationships are comparable to those reported for the original standardization sample of term infants, suggesting that the conceptional age corrections were appropriate (13).
In determining 6- to 12-month stability, both mental and psychomotor quotients at 24
weeks were more predictive of the psychomotor performance at 52 weeks than of mental
performance. It seems that psychomotor and mental quotients are better correlated with
each other at 6 months than at 1 year of age, and psychomotor scores at 6 months are
better predictors of mental and motor scores at 1 year of age than are 6-month mental
scores.
Prediction of developmental outcome
The significant predictor variables from the stepwise multiple regression of mental and
psychomotor quotients at 24 and 52 weeks are presented in Table 4. Only motor organization,
derived from the Brazelton examination at 4 weeks, and sleep-wake variables were significant predictors of both 24- and 52-week outcomes. Sleep-wake state variables predicted
Sleep, Vol. 8, No.3, 1985
T. F. ANDERS ET AL.
200
TABLE 2. Developmental course of sleep-wake
variables (Spearman correiations. rs)
Gestational age
Biological influences
QS%
Intercept
Slope
LSP
Intercept
Slope
TPI
Intercept
Slope
0.32
0.05
0.42"
-0.33
0.36b
-0.11
0.15
-0.06
0.36b
-0.20
0.3S b
-0.29
Biological and environmental influences
OOC%
- 0.45"
Intercept
0.44"
Slope
%LSP(l2-5 a.m.)
0.46"
Intercept
- 0.36b
Slope
Environmental influences
AS%
Intercept
Slope
AW%
Intercept
Slope
HTl
Intercept
Slope
Birth weight
-0.40"
-0.39 b
0.17
-O.OS
0.14
- 0.23
0.12
O.OS
-0.03
-0.12
-O.OS
0.09
-0.05
0.3S"
-0.05
0.16
Intercept. performance at birth; slope, developmental rate; QS%,
percentage of quiet sleep; LSP, longest sustained sleep period;
TPI, transition probability index; OOC%, percentage of sleep
time spent out of the crib; %LSP(l2-5 a.m.), proportion of LSP
occurring between 12-5 a.m.; AS%, percentage of active sleep;
AW%, percentage of time awake; HTI, holding time index.
'p ,,;:; 0.05.
bp,,;:;O.l.
more of the variance at 52 weeks of age than at 24 weeks of age, and more sleep-wake
state variables were significantly involved in prediction at the older age, even though all
of the variables were entered for each age.
Each of the predictor variables was also significantly correlated with other variables.
These relationships are summarized in Table 5. Some sleep-wake state variables that predicted developmental outcome (the holding time index, the AW%, and the LSP) were
correlated with non-motor Brazelton clusters (the state control dimension, the range of
state factor, and the summary factor), providing independent validation of their state relatedness. Sleep-wake state variables were also correlated with other sleep-wake parameters.
The motor behavior factor at 4 weeks did not correlate with any of the sleep-wake
variables. In contrast, the 4-week motoric processes dimension correlated with several
sleep-wake variables: the developmental rate of %LSP(l2-5 a.m.), the stability of the
OOC%, the stability of the AS%, and the birth status of the LSP. Because the motor
behavior factor correlated predominantly with other motor items, whereas the motor
Sleep, Vol. 8, No.3, 1985
SLEEP-WAKE STATES IN PREMATURE INFANTS
201
TABLE 3A. Bayley DQ
24 wks of age
52 wks of age
Psychomotor OQ
(SO) [range]
MentalOQ
(SO) [range]
98.4
(16.5)
[70-127]
93.0
(18.8)
[50-128J
104.8
(17.6)
[67-134]
99.3
(18.5)
[50-126J
Spearman correlations, rs
0.740.55 b
TABLE 3B. Stability of Bayley DQ at 2 ages
(Spearman correlations, rs)
24 wks of age
Psychomotor OQ
Mental OQ
0.730.67-
0.72-
52 wks of age
Psychomotor OQ
MentalOQ
0.47 e
OQ, development quotients.
-p ~ 0.001.
bp ~ 0.01.
ep ~ 0.05.
TABLE 4. Predictors of developmental outcome
(Stepwise Regression Analysis)
Step variable
Bayley mental OQ
24 wks of age
Motor behavior factor (4 wks)
rs
F
0.39
0.53
0.59-0.33
17.8 b
0.35
0.52
- 0.59- 0.40e
14.9b
7.6-
Bayley psychomotor OQ
24 wks of age
Motor behavior factor (4 wks)
OOC (DR)
0.31
0.46
0.42e
52 wks of age
Motor behavior factor (4 wks)
AW (BS)
OOC (DR)
0.21
0.31
0.52
0.21
0.35
HTI (DR)
52 wks of age
Motoric processes dimension (4 wks)
LSP (Stab)
0.53-
0.46e
5.g e
17.2b
8.610.39.39.1-
Stab, stability; DR, development rate; BS, birth status; r" Spearman correlations;
F, variance ratio. Other abbreviations as in Tables 2 and 3.
-p
bp
ep
~
~
~
0.01.
0.001.
0.05.
Sleep, Vol. 8, No.3, 1985
T. F. ANDERS ET AL.
202
TABLE 5. Outcome predictor variables and their correlates
Spearman correlations, rs
Sleep-wake variables
Brazelton items
Motor behavior factor (4 wks)
Summary factor score (4 wks)
Average factor summary
Summary dimension score (4 wks)
Motoric processes dimension
HTI (DR)
Range of state factor (4 wks)
State control dimension (4 wks)
AS (Stab)
OOC (DR)
AS
%LSP(l2-S a.m.)
LSP
TPI
Motoric processes dimension (4 wks)
Motor behavior factor (4 wks)
%LSP(l2-S a.m.)
Motoric processes dimension (2 wks)
ooe
AS
LSP
LSP (Stab)
TPI
Average summary factor
AW (BS)
DR
Stab
0.69a
O.SSb
- O.SSb
-0.S4b
O.SSb
- 0.42 e
- 0.40e
0.81a
0.63 b
0.S4b
O.SIb
- 0.S4b
- 0.S8 b
- O.Slb
-0.41e
- 0.44 e
-0.64a
- 0.42 e
0.S2 b
O.SOe
0.44 e
0.42 e
- 0.41e
O.SSb
- 0.46e
QS
TPI
AS
Range of state factor (2 wks)
BS
- 0.44c
-0.65 a
- 0.S6 b
- 0.47 e
0.45 c
0.49c
Abbreviations defined in Tables 2 and 4.
ap .:; 0.001.
bp ':; 0.01.
cp .:; O.OS.
dimension cluster correlated significantly with several sleep-wake variables, factor analysis
appears to be better than the a priori dimensions in distinguishing items reflecting state
organization from items reflecting solely motoric activity.
It is important to note in Table 5 that the predictor sleep-wake state components represent
developmental profiles, i.e., measures of birth status (intercept), developmental rate (slope),
and stability (root mean square error). In examining the relationships between sleep-wake
variables and predictor variables more than one component of a variable was, in some
cases, correlated with the predictor component. When two or three components were
significantly correlated with the predictor, all were listed.
Perinatal adversity
No significant relationship was demonstrated for obstetrical-perinatal risk status as a
predictor of outcome, although a trend was evident. The mean Bayley mental quotient in
the low-risk group (adversity scores ~6) was higher (106 vs. 101) than in a high-risk group
(adversity scores ~7) at 6 months. The same trend persisted at 1 year (101 vs. 95).
Sleep, Vol. 8, No.3, 1985
SLEEP-WAKE STATES IN PREMATURE INFANTS
203
Obstetrical-perinatal risk status was not significantly correlated with Brazelton variables
nor with any of the eight sleep-wake variables.
DISCUSSION
Many studies have demonstrated that neurobehavioral assessment of preterm infants is
complex. Due to the unstable nature of behavioral variables, single examinations are less
predictive of developmental outcome than repeated examinations, and examinations of
marginally impaired infants are less predictive than examinations of grossly impaired infants.
Stable and reliable early predictors of developmental outcome in risk infants are, nevertheless, important to establish.
Sleep studies during a single night, similarly, have been minimally useful in the search
for markers of damage or dysfunction. Polygraphic and electroencephalographic (EEG)
studies of sleep in the first days of life, when grossly abnormal, tend to predict follow-up
better than later polygraphic recordings (26,27). However, as with other tests, only severely
damaged infants demonstrate consistently abnormal sleep patterns or EEG wave forms
(C. Lombroso and Y. Matsumiya personal communication).
Yet, premature infants are known to be at risk for developmental delays, learning disabilities, and behavior problems----especially by the time they reach school age. Lags in
their development by 1 year of age are common, especially in the achievement of motor
milestones and size.
Our sample and their early life experiences are representative of most survivors of present
day premature nurseries, i.e., some infants under 1,000 gm, most under 2,000 gm. The
families of the infants were largely intact, educated, and advantaged so that, in this group
at least, the infants were not at double jeopardy. Despite their advantaged status, however,
the infants by 6 months of age demonstrated a range of outcomes on Bayley developmental
testing (67-134). By 1 year of age, the range was similar (50-128) although the average
level of performance had dropped slightly. Nevertheless, the developmental quotients, especially psychomotor developmental indices, were highly stable from 6 to 12 months of age.
In our study mental performance at 24 weeks was predicted by the developmental rate
of the holding time index while at 52 weeks it was predicted by the stability of the LSP.
At both ages, mental quotients were also predicted by Brazelton motor items.
These two sleep variables reflect temporal organization and continuity of sleep. Mature
holding time indices are characterized by long quiet sleep periods early in the night;
similarly, mature patterns of sustained sleep describe infants who sleep for long periods
uninterruptedly. Thus, infants whose sleep-wake state organization demonstrates mature
patterns of inhibitory activity are more likely to obtain high scores on the Bayley mental
scale. The negativity of the stability coefficient suggests that the infant's ability to be
consistent in sustaining long sleep episodes characterizes maturity. The positive correlation
between mental performance and the developmental rate of the holding time index can be
interpreted as signifying that infants who perform optimally at 6 months are infants whose
quiet sleep periods gradually and progressively lengthen during the first third of the night.
The sleep-wake state variables that reflect disrupted sleep and, indirectly, heightened
motor activity during the night, the developmental rate of time out of crib (OOC) and the
amount of wakefulness (AW) at birth, predicted Bayley psychomotor quotients. Infants who
had more wakefulness at birth, and who in the course of development showed gradual rates
of decline in their OOC time, performed better on the psychomotor scales. All components
of AS, which as a state is characterized by phasic motor activity, were significantly correlated
with the psychomotor predictors. The Brazelton motor behavior factor also predicted psySleep, Vol. 8, No.3, 1985
T. F. ANDERS ET AL.
204
_L .
"hl'lnll'ltor nerformance.
r -----
as it had nredicted mental performance. The motor behavior factor
includes measures of general tonus, motor maturity, activity level, the ability to pull-to-sit,
and defensive movements (15).
Are there then stable precursors of developmental outcomes? The present study points
to the domain of motor organization as a coherent and continuous system that merits further
study. The results demonstrate significant agreement between two independent approaches
to behavioral assessment of preterm infants, sleep-wake state organization and Brazelton
neonatal evaluation. Together these two distinct methods predict developmental outcome
at 6 months and 1 year of age.
All of the variables relate to the infant's activity during sleep and waking, and can be
viewed as measures of lack of inhibitory capacity or lack of motor regulation. In other
words, constructs such as activity, irregularity and dysinhibition-whether measured by
neonatal behavioral assessment or by sleep studies-are significant predictors of developmental outcome measured by the Bayley examination at 6 months and 1 year of age.
The results support the concept of continuity in a specific behavioral domain of development by suggesting that during the first year of life those indicators of development that
subserve motor organization are stable and predictive. Fukumoto et al. (28) studied the
developmental course of motor movements during sleep in premature and full-term infants
and found a stable and predictable course of maturation during the first year of life. They
suggested that motoric development followed a biologically determined course. Our results
extend those of Fukumoto et al. (28), by suggesting that an early stable determinant of
development is motor organization, that such organization is influenced by both biological
determinants and environmental experiences, and that the search for risk factors should
focus on the behavioral domain of motor regulation. Psychomotor performance and motor
organization may provide a foundation for studying subsequent continuities and transformations of motor behavior to cognitive and socio-emotional domains as they occur during
development. Longer term outcome cannot be predicted without a better understanding of
the protective and corrective environmental experiences that affect motor organization. An
awareness of early motor risk factors, however, may be useful in selecting particular infants
for continued observation and potential intervention.
Finally, what can be said about the maturation of sleep-wake states? It appears that the
course of development for QS, i. e., the infant's ability to regulate and sustain long periods
of inhibition, is to a large extent determined by the infant's biological maturity at birth.
The transition probability index, characterized by progressively more regulated state-tostate transitions, and the LSP also may follow this course. Maturation of these variables
seems to reflect central nervous system programming with minimal effects from environmental influences.
In contrast, the infant's experience with the environment interacts with its level of
biological development to affect the OOC and the capacity to entrain to the light-dark cycle
and sleep through the night, evidenced in the shift of long uninterrupted sleep periods to
the middle of the night. It is possible to speculate, for example, that a motorically immature
and irritable infant mayor may not be removed from the crib frequently at night depending
on the expectations and responses of the caregiver. These responses, in turn, will affect
the shift of sleep periods to the night and wakeful period to the day.
Wakefulness and the patterning of AS, in contrast, seem predominantly to be environmentally determined and not related to biological factors as judged by birth status. It is plausible
that, depending on the circumstances of the infant's environment, the two states reciprocally
.....,&.JL&....., .........
~
Sleep, Vol. 8, No.3, 1985
~..l
...
SLEEP-WAKE STATES IN PREMATURE INFANTS
205
shift in their proportional relationships in response to the more biological imperative of
QS. That is, infants whose environments provide the opportunity for more AS will have
less wakefulness while maintaining a stable and regular amount of QS and vice-versa. The
sparing of quiet sleep by the substitution of wakefulness or AS interchangeably, depending
on environmental circumstances, requires further research, but might account for the predominance of environmental influences which seem to affect the maturation of these states.
Since the sleep variables pursue courses of development that appear to reflect both
biological and environmental influences, it is interesting to speculate that these findings
support the studies that link Bayley outcomes to an interaction between biological maturity
and socio-emotional experiences occurring in early caregiving relationships (29). The concept that motor regularity and inhibition, and their opposite, irritability and activity, are
stable and characteristic of an infant, in both waking and sleeping behavior, is congruent
with the concept of both biological and environmental influences affecting the course of
development of sleep-wake state variables as well. Further research in the area of motor
organization, that is in the transformation from irritability to regulation, and in the relationship between the motor system and the cognitive, perceptual and socio-emotional systems
are warranted. Investigations of sleep-wake state maturation and organization, however,
continue to contribute significantly to this process, and to an understanding of early normal
and aberrant development.
Acknowledgment: This work was supported in part by the W T. Grant Foundation and the Irving
Harris Foundation. The paper was prepared while Dr. Anders was a Fellow at the Center for Advanced
Studies in the Behavioral Sciences and was supported in part by the John D. and Catherine T.
MacArthur Foundation.
REFERENCES
1. Field T, Dempsey J, Shuman H. Developmental follow-up of pre- and post-term infants. In: Friedman S,
Sigman M, eds. Preterm birth and psychological development. New York: Academic Press, 1981:299312.
2. Sigman M. Early development of pre term and full term infants: exploratory behavior in 8-month-olds.
Child Develop 1976;47:606--12.
3. Hunt J, Rhodes L. Mental development of preterm infants during the first year. Child Develop 1977;48:20410.
4. Sameroff A, Chandler M. Reproductive risk and the continuum of caretaking casualty. In: Horowitz F,
Scarr S, Siegel G, eds. Review of Child Development Research, Vol 4. Chicago: University of Chicago
Press, 1975;187-244.
5. Sostek A, Quinn P, Davitt M. Behavior development and neurologic status of premature and full-term
infants with varying medical complications. Field T, Sostek A, Goldberg S, Shulman H, eds. Infants Born
at Risk. New York: Spectrum, 1978;281-308.
6. Lipper E, Lee K, Gartner L, Grellong B. Determinants of neurobehavioral outcome in low birthweight
infants. Pediatrics 1981 ;67:502-5.
7. Hendricks J, Morrison A, Mann G. Different behaviors during paradoxical sleep without atonia depend
on pontine lesion site. Brain Research 1982;239:81-105.
8. Sterman MB, Hoppenbrouwers T. The development of sleep-waking and rest-activity patterns from fetus
to adult in man. In: Sterman M, McGinty D, Adinolfi A, eds. Brain development and behavior. New
York: Academic Press, 1971.
9. Ruckebusch Y, Barbey P. Les etats de sommeil chez Ie feotus et Ie nouveau-ne de las vache. CR Soc Bioi
1971 ;165: 1176--84.
10. Komer A, Zeanah C, Linden J, Berkowitz R, Agras S. The relation between neonatal and later activity
and temperament. Child Develop 1985;56:38-42.
11. Anders T, Keener M. The developmental course of nighttime sleep-wake patterns in full-term and premature
infants. Sleep 1985;8:173-92.
12. Brazelton TB. Neonatal Behavioral Assessment Scale. London: Spastic International Society, 1973.
13. Bayley N. Bayley Scales of Infant Development Manual. New York: Psychological Corporation, 1969.
14. Sostek A, Anders T. Relationships between the Brazelton neonatal scale, Bayley Infant Scales and early
temperament. Child Develop 1977;48:320--3.
Sleep, Vol. 8, No.3, 1985
206
T. F: ANDERSET Ak.
15. Lester B, Als H, Brazelton TB. Regional obstetrical anaesthesia and newborn behavior: a reanalysis toward
synergistic effects. Child Develop 1982;53:687-92.
16. Anders T, Sostek A. The use of time-lapse videorecording of sleep-wake behaviors in human infants.
Psychophysiology 1976;13:155-8.
17. Anders T, Keener M, Bowe T, Shoaff B. A longitudinal study of night time sleep-wake patterns in infants
from birth to one year. In: Call J, Galenson E, Tyson R, eds. Frontiers of Infant Psychiatry. New York:
Basic Books, 1983:150-66.
18. Fleiss J. Measuring nominal scale agreement among many raters. Psych Bull 1971;76:378-82.
19. Bowe T, Anders T. The use of the semi-Markov model in the study of the development of sleep-wake
states in infants. Psychophysiology 1979;16:41-8.
20. Bowe T. A systems analysis of the maturation of sleep using a semi-Markov model. [Dissertation]. Stanford
University, Stanford, California, 1981.
21. Siegel S. Nonparametric statistics for the behavioral sciences. New York: McGraw Hill, 1956.
22. Thoman E. A biological perspective and a behavioral model for assessment of premature infants. In: Bond
L, Joffee J, eds. Primary Prevention of Psychopathology. Vol 6: Facilitating Infant and Early Childhood
Development. Hanover, NH: Universities Press of New England, 1982: 159-79.
23. Kraemer H, Komer A, Hurwitz S. A model for assessing the development of preterm infants as a function
of gestational age, conceptional age or chronological age. Develop Psychology 1985 (in press).
24. Cox D. Analysis of binary data. London, Chapman Hall, 1977.
25. SAS Users Guide: Statistics, SAS Institute Inc. Carey, NC: SAS Institute Inc., 1982.
26. Monod N, E1iet-Flescher J, Dreyfus-Brisac C. Le sommeil du nouveau-ne du premature III. Les trouble
de I'organization du sommeil chez Ie nouveau-ne pathologique. Analyse d'etudes polygraphique. Bioi
Neonat 1967;11 :216-47.
27. Monod N, Pajot N, Guidasci S. The neonatal EEG: statistical studies and prognostic value in full-term
babies. Electroenceph Clin Neurophysiol 1972;32:529-44.
28. Fukumoto M, Mochizuki N, Takeishi M, Nomura Y, Segawa M. Studies of body movements during night
sleep in infancy. Brain Dev 1981;3:37-43.
29. Crnic K, Ragozin A, Greenberg M, Robinson N, Bashan R. Social interaction and developmental competence of preterm and full term infants during the first year of life. Child Develop 1983;54:1199-210.
Sleep, Vol. 8, No.3, 1985