Van Cauter E, Leproult R, Plat L. Age-Related Changes in Slow Wave Sleep and REM Sleep and Relationship With Growth Hormone and Cortisol Levels in Healthy Men. JAMA. 2000;284(7):861–868. doi:10.1001/jama.284.7.861
Author Affiliations: Department of Medicine, University of Chicago, Chicago, Ill.
Context In young adults, sleep affects the regulation of growth hormone (GH)
and cortisol. The relationship between decreased sleep quality in older adults
and age-related changes in the regulation of GH and cortisol is unknown.
Objective To determine the chronology of age-related changes in sleep duration
and quality (sleep stages) in healthy men and whether concomitant alterations
occur in GH and cortisol levels.
Design and Setting Data combined from a series of studies conducted between 1985 and 1999
at 4 laboratories.
Subjects A total of 149 healthy men, aged 16 to 83 years, with a mean (SD) body
mass index of 24.1 (2.3) kg/m2, without sleep complaints or histories
of endocrine, psychiatric, or sleep disorders.
Main Outcome Measures Twenty-four–hour profiles of plasma GH and cortisol levels and
polygraphic sleep recordings.
Results The mean (SEM) percentage of deep slow wave sleep decreased from 18.9%
(1.3%) during early adulthood (age 16-25 years) to 3.4% (1.0%) during midlife
(age 36-50 years) and was replaced by lighter sleep (stages 1 and 2) without
significant increases in sleep fragmentation or decreases in rapid eye movement
(REM) sleep. The transition from midlife to late life (age 71-83 years) involved
no further significant decrease in slow wave sleep but an increase in time
awake of 28 minutes per decade at the expense of decreases in both light non-REM
sleep (−24 minutes per decade; P<.001) and
REM sleep (−10 minutes per decade; P<.001).
The decline in slow wave sleep from early adulthood to midlife was paralleled
by a major decline in GH secretion (−372 µg per decade; P<.001). From midlife to late life, GH secretion further
declined at a slower rate (−43 µg per decade; P<.02). Independently of age, the amount of GH secretion was significantly
associated with slow wave sleep (P<.001). Increasing
age was associated with an elevation of evening cortisol levels (+19.3 nmol/L
per decade; P<.001) that became significant only
after age 50 years, when sleep became more fragmented and REM sleep declined.
A trend for an association between lower amounts of REM sleep and higher evening
cortisol concentrations independent of age was detected (P<.10).
Conclusions In men, age-related changes in slow wave sleep and REM sleep occur with
markedly different chronologies and are each associated with specific hormonal
alterations. Future studies should evaluate whether strategies to enhance
sleep quality may have beneficial hormonal effects.
Decreased subjective sleep quality is one of the most common health
complaints of older adults.1 The most consistent
alterations associated with normal aging include increased number and duration
of awakenings and decreased amounts of deep slow wave (SW) sleep (ie, stages
3 and 4 of non–rapid eye movement (non-REM) sleep).2- 4
REM sleep appears to be relatively better preserved during aging.3- 7
The age at which changes in amount and distribution of sleep stages appear
is unclear because the majority of studies have been based on comparisons
of young vs older adults. Several investigators have noticed that there are
marked decreases in SW sleep in early adulthood in men but not in women.8- 11
Sleep is a major modulator of endocrine function, particularly of pituitary-dependent
hormonal release. Growth hormone (GH) secretion is stimulated during sleep
and, in men, 60% to 70% of daily GH secretion occurs during early sleep, in
association with SW sleep.12 Whether decrements
in SW sleep contribute to the well-known decrease in GH secretion in normal
aging is not known.13- 15
In contrast to the enhanced activity of the GH axis during sleep, the
hypothalamic-pituitary-adrenal (HPA) axis is acutely inhibited during early
SW sleep.16- 20
Furthermore, even partial sleep deprivation results in an elevation of cortisol
levels the following evening.21 Thus, both
decreased SW sleep and sleep loss resulting from increased sleep fragmentation
could contribute to elevating cortisol levels. An elevation of evening cortisol
levels is a hallmark of aging14,15,22
that is thought to reflect an impairment of the negative feedback control
of the HPA axis and could underlie a constellation of metabolic and cognitive
The present study defines the chronology of age-related changes in sleep
duration and quality (ie, amounts of sleep stages), GH secretion, and cortisol
levels in healthy men and examines whether decrements in sleep quality are
associated with alterations of GH and cortisol levels.
Data from a total of 149 healthy men, aged 16 to 83 years, are presented.
Mean (SD) body mass index (BMI) of the subjects was 24.1 (2.3) kg/m2. All subjects were of normal weight (BMI,18-28 kg/m2).
The data were collected between 1985 and 1999 in a series of studies
from our group (109 out of 149 individual data sets14,26- 32)
and 3 other laboratories using similar assay procedures, recording procedures,
or both (University of Pittsburgh, Pittsburgh, Pa, 14 subjects33,34;
University of California, Los Angeles, 8 subjects35,36;
and Pennsylvania State University, Hershey, 18 subjects37).
Except for data from 29 subjects studied in our laboratory, all other data
were included in previously published reports that did not address the chronology
of age effects. The raw data from all 149 subjects were submitted to a new
analysis designed to quantitatively define sleep and hormonal parameters across
The subjects were paid volunteers who had no sleep complaints, did not
take any drugs, were in good health based on a physical examination, and had
no history of endocrine, psychiatric, or sleep disorders. Shift workers, subjects
recently having traveled across time zones, and competitive athletes were
excluded. The young subjects had reached Tanner stage 5 of sexual maturity
and full body height. The older men were self-sufficient and had no cognitive
impairment. All subjects gave written informed consent. The experimental protocol
was approved by the ethics review board at each institution.
Prior to the study, the subjects spent 1 to 3 habituation nights in
the sleep laboratory. In studies including hormonal measurements (133 of 149
studies), a catheter was inserted into a forearm vein and blood samples were
collected at 15- to 30-minute intervals for 24 to 25 hours. During the night,
the catheter was connected to tubing extending to an adjacent room to avoid
disturbing the subject. All-night polygraphic sleep recordings were obtained.
The subjects remained recumbent in bed in darkness for at least 8 hours. Daytime
naps were not allowed.
Sleep, cortisol, and GH profiles were obtained in 132, 124, and 114
subjects, respectively. Concomitant sleep, cortisol, and GH profiles were
obtained in 94 subjects.
Polygraphic sleep recordings were visually scored at 20- or 30-second
intervals in stages wake, 1, 2, 3, and 4, and REM using standardized criteria.38 Sleep onset and morning awakening were defined, respectively,
as the times of occurrence of the first and last interval scored as stage
2, 3, 4, or REM. The sleep period was defined as the interval separating sleep
onset from morning awakening. The total sleep time was calculated as the sleep
period minus the total duration of awakenings. The total duration of each
stage was expressed in minutes as well as a percentage of the sleep period.
Slow wave sleep was defined as the sum of stages 3 and 4.
In all studies, cortisol levels were measured by a standard radioimmunoassay
(RIA). The limit of sensitivity averaged 27.6 nmol/L and intra-assay coefficients
of variation (CV) were 5% to 10%.
In 89 profiles, GH concentrations were measured by an RIA with a sensitivity
of 0.4 µg/L.39 Intra-assay CV ranged
from 5% to 9%. The interassay CV averaged 15%. In 25 studies, GH concentrations
were measured by a chemiluminescence method (8 studies: Nichols Institute
Diagnostics, San Juan Capistrano, Calif; and 17 studies: Diagnostic Product
Corporation, Los Angeles, Calif) with a limit of sensitivity of 0.002 to 0.003
µg/L, an intra-assay CV ranging from 4.8% to 9.9%, and an interassay
CV less than 8%. Baseline, ie, nonpulsatile, concentrations of GH less than
1 µg/L by chemiluminescence correspond to concentrations less than the
limit of sensitivity (0.4 µg/L) by RIA. Estimations of pulsatile GH
secretion greater than baseline levels derived from GH profiles measured by
chemiluminescence do not differ significantly from those measured by RIA.40
The circadian variation of plasma cortisol was quantified using a best-fit
curve based on periodogram calculations.41
The acrophase and nadir were defined, respectively, as the times of occurrence
of the maximum and minimum of the best-fit curve. The value of the acrophase
or nadir was defined as the level attained by the best-fit curve at the acrophase
Significant pulses of GH secretion greater than baseline levels were
identified using a computerized algorithm.42
The threshold for significance was set at 2 times the intra-assay CV. For
each significant pulse, the amount of GH secreted above baseline level was
estimated by mathematical deconvolution based on a 1-compartment model for
GH clearance and variable individual half-lives.43
The total amount of GH secreted over a given time interval was determined
by summing the amounts secreted in each of the pulses occurring during that
Each of the parameters used to quantify sleep, GH secretion, and the
24-hour cortisol profile was used in 2 analyses. First, the parameter was
considered as a dependent variable in an analysis of variance (ANOVA) including
age, BMI, and the interaction age × BMI as independent variables. When
the interaction age × BMI had P>.05 based on
type III sums of squares, the analysis was repeated with only age and BMI
as independent variables. To verify that significant interactions did not
reflect the impact of a single subject, the calculations were repeated after
excluding data from the most outlying subject for the variables in the analysis.
The interaction was maintained in the analysis only if the statistical significance
was not critically dependent on a single subject. Second, the data were grouped
by age ranges (aged 16-25 years, 42 subjects; aged 26-35 years, 28 subjects;
aged 36-50 years: 26 subjects; aged 51-60 years, 23 subjects; aged 61-70 years,
18 subjects; and aged 71-83 years, 12 subjects). For each parameter, simple
linear regressions with age as an independent variable were calculated separately
for subjects from early adulthood (aged 16-25 years) to 43 years, ie, the
midpoint of the midlife range (aged 36-50 years), and for subjects from midlife
to late life (aged 44-83 years). Unless otherwise indicated, all group values
are expressed as mean (SEM).
Consistent with previous reports,3- 5
the sleep period was not significantly affected by age. In contrast, total
sleep time decreased markedly with aging (P<.001),
but significant reductions in total sleep time did not occur until after midlife.
From midlife until the eighth decade, total sleep time decreased, on average,
by 27 minute per decade (Table 1).
Aging had a differential impact on sleep parameters (Figure 1). From early adulthood to midlife (age 16-25 to 36-50 years),
the percentage of SW sleep decreased from 18.9% (1.3%) to 3.4% (1.0%), and
this decrease in deep non-REM sleep was compensated by an increase in light
non-REM sleep (ie, stages 1 and 2) from 51.2% (1.4 %) to 67.3% (1.6 %) without
significant change in time spent awake. There were no changes in REM sleep
from early adulthood to midlife. Increases in wake time and decreases in REM
sleep became significant starting at midlife, and stages 1 and 2 decreased
from 60.5% (2.5%) for subjects aged 51 to 60 years to 50.6% (4.6%) for subjects
older than 70 years. Changes in SW sleep after age 50 years were not significant.
Effects of BMI and of the interaction age × BMI were significant
for stages 1 and 2 and SW sleep, but not for other sleep parameters (Figure 1). In young to middle-aged subjects
(aged 16-43 years), but not in older adults (aged 44-83 years), higher BMI
was associated with shallower, non-REM sleep (less SW sleep, more stages 1
and 2 sleep).
Mean 24-hour GH profiles from 8 older men and 8 young men who were matched
for BMI illustrate the temporal coincidence of the major GH pulse with early
sleep and the marked reduction in GH levels in old age (Figure 2).
The impact of age on GH secretion during the 24-hour cycle, wake time,
and sleep, is illustrated in Figure 3.
Significant effects of age independent of BMI were evident. From young adulthood
to midlife, GH secretion decreased by nearly 75%. Further smaller decreases
occurred between midlife and late adulthood (Table 1).
A significant negative association of BMI with 24-hour GH secretion
and GH secretion during waking, but not during sleep, was detected independently
Mean 24-hour cortisol profiles in young and older men are shown in the
lower panels of Figure 2. In both
groups, cortisol levels show an early morning elevation, declining levels
throughout the daytime, and a nocturnal quiescent period. Age differences
are mostly apparent in the evening and early part of the night.
Figure 4 illustrates the changes
in 24-hour mean cortisol level, morning acrophase, and evening nadir across
adulthood. A modest effect of aging on the 24-hour mean cortisol level was
detected. Aging was associated with an elevation of the evening nadir, but
morning maximum values remained stable across all age ranges. Increases in
evening cortisol levels became apparent after midlife (Table 1).
There were no effects of BMI or of the interaction BMI × age on
any of the parameters characterizing the 24-hour cortisol profile.
Age-related decreases in GH secretion and SW sleep followed a similar
chronology, with the majority of the decrements occurring in young adulthood,
whereas age-related increases in evening cortisol did not occur until the
fifth decade, when decreases in REM sleep and increases in amount of wake
time became apparent (Table 1).
We thus sought to determine whether the concomitant sleep alteration and its
interaction with age contributed to the hormonal changes.
Analysis of variance of GH secretion during sleep in relation to age,
SW sleep, and their interaction indicated that SW sleep (P<.001) and the interaction age × SW sleep (P = .008) accounted for the majority of the variance and that effects
of age per se were nonsignificant (P>.50). Similar
findings were obtained when total 24-hour GH secretion was analyzed (age, P>.50; SW sleep, <.001; age
× SW sleep, P = .003). In young to middle-aged
subjects, but not in older men, increased amounts of SW sleep were associated
with higher levels of GH secretion. The left panels of Figure 5 compare GH levels during sleep in subjects who had large
amounts of SW sleep and in age- and BMI-matched subjects who had small amounts
of SW sleep.
The variance of evening cortisol levels was analyzed in relation to
age, REM sleep, wake time, and their interactions. The contributions of all
interactions and of wake time were not significant. Age (P<.001) and, to a lesser extent, REM sleep (P<.10) were both negatively related to evening cortisol concentrations.
To illustrate the inverse relationship between amounts of REM sleep and evening
cortisol levels, the right panels of Figure
5 show the mean cortisol nadir in subjects with large amounts of
REM sleep and in age- and BMI-matched subjects with small amounts of REM sleep.
The present analysis demonstrates that, in healthy men, aging affects
SW sleep and GH release with a similar chronology characterized by major decrements
from early adulthood to midlife. In contrast, the impact of age on REM sleep,
sleep fragmentation, and HPA function does not become apparent until later
in life. The analysis further suggests that age-related alterations in the
somatotropic and corticotropic axes may partially reflect decreased sleep
Human sleep is under the dual control of circadian rhythmicity and of
a homeostatic process relating the depth of sleep to the duration of prior
wakefulness.44 This homeostatic process involves
a putative neural sleep factor that increases during waking and decays exponentially
during sleep. Slow wave sleep is primarily controlled by the homeostatic process.
Circadian rhythmicity is an oscillation with a near 24-hour period generated
by a pacemaker located in the hypothalamic suprachiasmatic nucleus. Circadian
rhythmicity plays an important role in sleep timing, sleep consolidation,
and the distribution of REM sleep.45 The present
data indicate that an alteration in sleep-wake homeostasis is an early biological
marker of aging in adult men. In contrast, components of sleep that are under
the control of the circadian pacemaker appear to be relatively well preserved
until late in life.
The chronology of aging of GH secretion follows a pattern remarkably
similar to that of SW sleep. Thus, in men, the so-called "somatopause" occurs
early in adulthood, between age 25 and 35 years, an age range that corresponds
to the human life expectancy before the development of modern civilization
and is essentially completed by the end of the fourth decade. Our analyses
further indicate that reduced amounts of SW sleep, independent of age, are
partly responsible for reduced GH secretion in midlife and late life. That
this correlative evidence reflects a common mechanism underlying SW sleep
generation and GH release rather than an indirect association is supported
by 2 studies that have shown that pharmacological enhancement of SW sleep
results in increased GH release.46,47
Also supporting a causal relationship between decreased sleep quality and
reduced nocturnal GH secretion are studies in patients with sleep apnea showing
a marked increase in GH release following treatment with positive airway pressure.48,49 The reverse interaction between sleep
and GH, ie, a deleterious impact of reduced somatotropic function on sleep,
is also possible since studies in both normal and pathological conditions
have shown that GH-releasing factor and GH influence sleep quality.12,50 In the present study of nonobese
men, the finding of a negative impact of BMI on both GH secretion during waking
and amount of SW sleep is consistent with the hypothesis that inhibition of
the GH axis may adversely affect sleep regulation.
While the clinical implications of decreased SW sleep are still unclear,
the relative GH deficiency of the elderly is associated with increased fat
tissue and abdominal obesity, reduced muscle mass and strength, and reduced
exercise capacity.51- 53
Multiple trials are currently examining the clinical usefulness and safety
of replacement therapy with recombinant GH, the other hormones of the GH axis,
and synthetic GH secretagogues in elderly adults without pathological GH deficiency.
While the benefits of such interventions are still unproven, the present findings
suggest that they should target a younger age range than currently envisioned,
ie, individuals in early midlife rather than those older than 65 years, when
peripheral tissues have been continuously exposed to very low levels of GH
for at least 2 decades. Furthermore, since pharmacological enhancement of
SW sleep in young adults has been shown to result in a simultaneous and proportional
increase in GH release46,47 and
ongoing studies in our laboratory indicate that similar effects can be obtained
in older subjects, drugs that reliably stimulate SW sleep may represent a
novel class of GH secretagogues.
The present data demonstrate that the amount of REM sleep is reduced
by approximately 50% in late life vs young adulthood. However, reduced amounts
of REM sleep and significant sleep fragmentation do not occur until after
age 50 years. The impact of aging on cortisol levels followed the same chronology.
Aging was associated with an elevation of evening cortisol levels, reflecting
an impaired ability to achieve evening quiescence following morning stimulation.
Studies in both animals and humans have indicated that deleterious effects
of HPA hyperactivity are more pronounced at the time of the trough of the
rhythm than at the time of the peak.25,54
Thus, modest elevations in evening cortisol levels could facilitate the development
of central and peripheral disturbances associated with glucocorticoid excess,
such as memory deficits and insulin resistance,24,25
and further promote sleep fragmentation. Indeed, elevated cortisol levels
may promote awakenings.55,56
Elevated evening cortisol levels in late life probably reflect an impairment
of the negative feedback control of the HPA axis in aging. Our analyses suggest
that there is a relationship between this alteration of HPA function and decreased
amounts of REM sleep that is independent of age. The data generally support
the concept that decreased sleep quality contributes to the allostatic load,
ie, the wear and tear resulting from overactivity of stress-responsive systems.57
The present study focused on the effects of aging on the relationship
between sleep and the somatotopic and corticotropic axes in men because the
predominant GH secretion occurs during sleep in men but not in women11 and because there is evidence to suggest that the
marked decreases in SW sleep in early adulthood occur in men but not in women.8- 11 Whether
conclusions similar to those obtained for men hold for women will require
a separate evaluation as sex differences in sleep quality as well as 24-hour
profiles of GH and cortisol secretion have been well documented in both young
and older adults.11,12,22
In conclusion, in healthy men, the distinct changes in sleep quality
that characterize the transitions from early adulthood to midlife, on the
one hand, and from midlife to old age, on the other hand, are each associated
with specific alterations in hormonal systems that are essential for metabolic
regulation. Strategies to prevent or limit decrements of sleep quality in
midlife and late life may therefore represent an indirect form of hormonal
therapy with possible beneficial health consequences.