By PATRICIA PRINZl, SANDRA BAILEyl, KAREN MOE2, CHARLES WILKINSON2 and JAMES SCANLAN2
 

Cortisol and sleep in health

There is growing literature linking sleep and endocrine system responsiveness to stress. The sleep disruptive effects of stress are well known, and are likely mediated via CRH release (Opp 1995), which in turn stimulates endogenous ACTH and cortisol secretion. Disruption in the sleep cycle may in itself act as a stress, and be reflected in hypothalamicpituitary-adrenal (HP A) axis activation. In healthy adult males, total or partial sleep deprivation was followed by increased cortisol levels the following day (Leproult et al. 1997). Elevated mean 24-h cortisol levels have been observed in habitually poor as compared with good sleepers (Madjirova et al. 1995).

The inverse relationship between endogenous cortisol and sleep does not remain with exogenously administered glucocorticoids, which enhance NREM sleep (Steiger et al. 1998; Bohlhalter et al. 1997; Born et al. 1989; Friess et al. 1995).

Exogenous cortisol may not be an appropriate model for study of endogenous cortisol-sleep interactions, because exogenous cortisol may increase sleep as a result of increased CRH inhibition (Opp 1995; Chang and Opp 1998; Tsuchiyama et al. 1995). It is well established that an earlier rise in endogenous cortisol is associated with an earlier morning awakening (Van Cauter et al. 1998). In a recent study of endogenous cortiso1sleep rhythms, lower cortisol secretion rates were temporally associated with increased slow wave sleep (SWS) bouts (Gronfier et al. 1999), lending support for the view that low, rather than high, levels of endogenous cortisol at the circadian nadir foster better sleep.

Aging, cortisol and sleep

Less is known is about the relationship between cortisol and sleep in older adults. There are marked sleep changes exhibited with aging. Older adults have less SWS and enhanced fragmentation of sleep by wakefulness (Prinz 1995). Adults are also known to become more 'morning' as they age, with earlier retiring and arising times (Van Cauter et al. 1996) and an earlier timing of morning cortisol secretion (Van Cauter et al. 1998; van Coevorden et al. 1991). Healthy older adults have elevated cortiso11evels at the time of the circadian nadir (Van Cauter et al. 1996) and have higher basal cortisol levels than younger adults (Belanger et al. 1994, Copinschi and Van Cauter 1995; Van Cauter et al. 1996; Deuschle et al. 1997, Nicolson et al. 1997). The circadian amplitude (range of values from highest to lowest) of many variables, including cortisol (Deuschle et al. 1997; Van Cauter et al. 1996) and temperature (Vitiello et al. 1986, Weitzman et al. 1982), decreases with age in both women and men.

Gender differences in cortisol, sleep and stress responsivity.

In adults, cortisol levels are lower in premenopausal females than in males of similar ages (Van Cauter et al. 1996; Bohlhalter et al. 1997; Madjirova et al. 1995). After menopause, there are no longer discernible gender differences in overall total plasma cortisol levels. It is of considerable interest that the level of the circadian acrophase of cortisol may be age elevated in women, but not in men (Van Cauter et al. 1996). Gender differences in stress responsiveness have been hypothesized. Older women, when subjected to a mild stress, show greater effects in sleep parameters than age-matched men. These sleep disruptions in women were manifested by increased sleep latency, decreased time in bed, decreased total sleep time, and increased total wake time (Vitiello et al. 1996).

Older women exhibit a slightly greater cortisol response to a simulated challenge than do age-matched men (Nicolson et al. 1997; Seeman et al. 1995), which may reflect declining estrogen in older women. There is growing evidence indicating that estrogen supplementation may ameliorate many of the effects of stress (Kirschbaum et al. 1995, Lindheim et al. 1992, Saab et al. 1989, von Eiff et al. 1971). Komesaroff et al. 1999 observed that blood pressure, glucocorticoid and catecholamine responses to psychological stress were reduced by estrogen supplementation in perimenopausal women. Similar findings were reported for postmenopausal women (Del Rio et al. 1998, Lindheim et al. 1992), and also for female primates (Kaplan et al. 1996).

We hypothesized that estrogen replacement therapy (ERT) status might also ameliorate age increases in cortisol levels seen during sleep. The studies of cortisol-sleep relationships that are reviewed above uniformly reported plasma cortisol data as the total plasma level, which includes bound as well as unbound cortisol. The confounding influence of binding proteins can be sizeable. Plasma binding proteins including cortisol binding globulin (CBG) and albumin can vary in response to stress (Fleshner et al. 1995, Tannenbaum et al. 1997) and/or circadian factors (Meaney et al. 1995), as well as gender (Deuschle et al. 1996, Stolk et al. 1996) and estrogen therapy (Darj et al. 1993). All of these factors can mask cortisol-sleep relationships when total rather than free cortisol data are used. For this reason, recent research focuses on urinary or salivary free cortisol assessments: 24-h urinary free cortisol (UFC) levels can index hypercortisolemia and its functional correlates (Litchfield et al. 1998).

Here we report a study in which 24-h UFC (an index of 24-h HPA activation), was used to examine interactions among cortisol, sleep, and sleep timing in healthy senior women, half of whom were on ERT. The women were studied under baseline conditions and a mildly stressful condition, a 24-h indwelling IV catheter that has been shown to affect sleep adversely in this population (Vitiello et al. 1996), so that stress reactivity effects on sleep and cortisol could also be assessed in relation to Estrogen Replacement Therapy status.

METHODS
Subjects and recruitment

Older subjects (> 55 years) were recruited from western Washington to participate in a large cross-sectional study on hormones, sleep and cognition in normal aging (AG-12915). Public service announcements in local newsletters, as well as letters to senior centers, retirement communities, and groups of retirees were used to elicit initial interest. Recruitment posters emphasized the requirement for healthy nonsmokers with no memory impairment, obesity, or sleep disorders. Individuals who responded were screened via a three-step protocol to ensure that all study subjects met the study inclusion/exclusion criteria.

Of potential subjects, 43% passed the first screening, a 30-minute telephone interview focusing on general medical history, use of tobacco and medications, and sleep. A clinical interview and physical examination with health history followed; 63% of these subjects met inclusion/exclusion criteria (Table 1). Information about menopause and hormone replacement therapy, if any, was also obtained. Cognitive status (Folstein Mini-Mental Status Exam [MMSE]) (Folstein et al. 1975), depression (Center for Epidemiological Studies?.

Exclusion criteria

Age less than 55 years
Major physical illness requiring treatment in previous year
Currently taking any medication with CNS effects, including but not limited to antidepressants, beta blockers, barbiturates, benzodiazepines, sedatives
Current or past (last 6 months) use of any steroids other than estrogen
or progesterone
Significant stress or anxiety in past 6 months
Current or past psychiatric illness, including depression
Uncorrected thyroid problem (4.6 I1g dL -1 < FT41 < 10.8 I1g dL -1) Vegan or other very restricted diet
Body Mass Index < 18 or > 33
Tobacco use
Sleep disorders
Pacemaker
Uncontrolled hypertension
Mini - Mental Status score < 25
History of:
Neurological disorders
Chronic renal, hepatic, or pulmonary disease Insulin-dependent diabetes
Other endocrine disorders (except controlled thyroid) Significant loss of consciousness
ETOH or drug abuse

Depression Scale [CES-D]) (Radloff 1977), and anxiety (SCL90 Anxiety Scale) (Derogatis et al. 1976) were assessed during the interview. In addition, all subjects completed the Pittsburgh Sleep Quality Index, geriatric version (PSQI) (Buysse et al. 1991).

The University of Washington Human Subject Committee approved the study protocol. Informed written consent was obtained from all participants. Presented here are data from a subset of 42 female subjects (age = 69.6 ± 6.2 years), who participated in that study's protocol for whom baseline and stressed UFC values were available. Of the subset of women selected for this study, 20 were on long-term Estrogen Replacement Therapy and 22 were not (Moe et al. 1998). Of the 20 women on ERT, seven were also taking oral progesterone.

General procedures

Subjects were admitted to the General Clinical Research Center (GCRC) at the University of Washington Medical Center, from early Tuesday evening through Friday morning. While there, they were encouraged to adhere to their customary bed and risetimes, and were discouraged from napping. Subjects were strongly encouraged to remain as active as possible during daytime hours (e.g. walking around the unit). All meals were provided through the GCRC Nutrition Research Kitchen, with a daily dietary composition of 52% carbohydrate, 18% protein, and 30% fat. Total daily calories were based on age, height and weight. All voided urine was collected across the last two 24-h periods for UFC determination. Subjects retired at their customary bedtimes and slept until their customary risetime or until spontaneous awakening, and arising and bedtimes were recorded.

Plasma sampling

A forearm indwelling iv catheter (iv cath) was introduced at 08.00 h at the completion of the baseline (non iv) 24-h period. Blood sampling through the iv continued every 20 min for 24 h (the stress condition). Less than 300 mL total blood was drawn from each subject. Between sampling intervals, a heparinized (2000 U L-1) iv saline solution was infused at a slow rate to maintain catheter patency. Night-time samples were drawn remotely from an adjoining room by use of extension iv tubing. The study was completed at the end of the blood-sampling period, 08.40 h the next morning.

Sleep and EEG recording

The first night was an adaptation and apnea screening night. None of the subjects reported here showed evidence of significant levels of sleep apnea. Subjects underwent a series of cognitive tests on the following day. Sleep EEG was recorded on Nights 2 and 3.
Standard procedures (Rechtschaffen and Kales 1968) were used for sleep recordings, including electroencephalogram (EEG), electro-oculogram (EOG), and electromyogram (EMG). EEG electrodes were positioned for conventional sleep recordings at C3, C4, 01, and 02 (international 10-20 system of measurement) and were referenced to the contralateral mastoids. Data were recorded using a Grass 8-24 polygraph with filter settings of 0.1 and 70 Hz. Four channels (C3, C4, EMG, 02) were digitized simultaneously using a 12-bit digitizer installed in a microcomputer at a voltage range of ± 2.5 V. The sampling rate was 256 Hz, with data averaging and decimation to 128 Hz. All recordings were calibration corrected. Additional recording details are reported elsewhere (Prinz et al. 1994).

Sleep scoring

Paper records were scored for sleep stages by a reliable (r = 0.86, P < 0.01) and highly trained human rater using standard technique (Rechtschaffen and Kales 1968). Standard polysomnographic variables were then calculated. All sleepwake variables reported here are from the human-scored data, except for measures of stages 3 + 4 sleep (slow wave sleep, SWS). Minutes of SWS were calculated by the most recent version of the C STAGE algorithm that our laboratory developed to quantify and score EEG epochs using EEG spectral analysis (Prinz et al. 1994). For older populations, SWS values based on this algorithm are less subject to scorer bias and individual differences in overall spectral energy (Prinz et al. 1994; Larsen et al. 1995).

Urinary free cortisol

From the 24-h collected urine, samples were extracted by mixing thoroughly with chilled methylene chloride, removing an aliquot of the organic phase, and drying. Dried samples were reconstituted with buffer and assayed as previously described for plasma cortisol (Wilkinson et al. 1997).

Estrogen

A pooled 24-h blood sample was collected from each subject for determination of plasma estrogen levels. Total estrogens (estrone + estradiol) were measured by using a solvent extraction followed by column chromatography and radioimmunoassay.

Data Analysis

Unless otherwise stated, data are reported as means ± standard deviation (SD). All variables were examined prior to analyses for extreme (> 3 SD from the mean) outliers and normality of distributions. Significantly non-normal distributions were transformed prior to analyses or were analyzed using non parametric tests. UFC values and most standard polysomnographic sleep variables were examined using multiple regressions (SPSS-PC statistical software, SPSS Inc, Chicago, IL, USA) that controlled for the effects of age. Statistical significance was set al P 0.05 (two-tailed).

RESULTS

All 42 healthy subjects were found to score in the healthy normal range for depression, anxiety, and cognitive abilities. Sleep data from one subject were not available for the second (stress) night; for that night the total sample size was 41. As can be seen in Table 2, the stress condition (a 24-h iv catheter) significantly impaired sleep efficiency, minutes of nonrem sleep (stages 2, 3 + 4) and minutes of REM, and increased wakefulness (wake after sleep onset; W ASO) and total numbers of awakenings lasting I or more min (TA WI). In addition, rise time was earlier and UFC was higher in the stress condition as compared with baseline levels. Group differences (ER T vs. non ERT) were generally nonsignificant except for estrogen levels (Table 2). In a repeat analysis (not shown), the seven women on progesterone-opposed Estrogen Replacement Therapy were removed from the Estrogen Replacement Therapy group; several variables now differed significantly in women on unopposed Estrogen Replacement Therapy as compared with non ER T women (all P < 0.05): minutes of stages 2,3,4 sleep and risetime were greater, and cortisol was lower, all for the stress condition.

Table 3 describes UFC-sleep correlations for the combined groups of the 41 women (sleep data from one non Estrogen Replacement Therapy subject were not available for the stressed night). Higher baseline UFC levels were associated with earlier time of habitual rising as reported on the PSQI, earlier risetime on the stressed night, and reduced REM sleep on both the baseline and stressed nights (P = 0.03-0.001) (Table 3). Table 3 also shows residual statistics formed when baseline values are regressed on stress condition values, to form an indication of the stress condition value as adjusted for baseline levels. Correlations for residuals were similar to those of the stress condition.
 
Higher UFC levels in the stress condition were also associated with sleep impairment, with profiles of lower REM sleep, and earlier risetimes (each on the stress night) similar to those of baseline cortisol. In addition, higher UFC was associated with lower sleep efficiency and decreased minutes of stages 2, 3 and 4 sleep in the stressed, but not the baseline condition. Ten to 21 percent of the variance in these measures was attributable to UFC levels in the stressed condition.

When UFC levels attributable to stress were examined separately (as the residual of U.FC in the stress condition, controlling for baseline UFC), this measure correlated negatively with NREM sleep measures, but not with REM sleep (Table 3). Overall, Table 3 shows a pattern of many significant negative correlations and no positive correlations between UFC and sleep. In parallel analyses using a UFC to creatinine ratio instead of 24-h UFC, similar but weaker associations with REM sleep and with risetime were observed; the associations with NREM sleep were nonsignificant.
UFC-sleep relationships were also examined separately for the Estrogen Replacement Therapy vs. non Estrogen Replacement Therapy groups. As seen in Table 4, a similar pattern that included numerous significant negative UFC-sleep relationships emerged for the non ER T group; significant UFC-sleep correlations were absent among Estrogen Replacement Therapy women.

While estrogen and UFC for the baseline condition were not related, estrogen was negatively associated with UFC levels during the stressed conditions (r = -0.31, d.f. = 34, P = 0.06) and with the residual ofUFC (stressed night UFC adjusted for baseline night UFC) (r = -0.35, d.f. = 33, P = 0.04). In separate group analyses, similar UFC-estrogen correlationships were observed in the non Estrogen Replacement Therapy (but not ERT) group.

DISCUSSION

The purpose of this study was to examine the relationship between UFC and sleep quality under both baseline and mildly stressful conditions in healthy older women. The higher UFC levels in the mild stress condition for the non Estrogen Replacement Therapy women were consistently associated with sleep impairment (i.e. lowered sleep efficiency; less time in sleep stages 2, 3 and 4; fewer minutes of REM sleep; and earlier risetime). In most (Follenius et al. 1992, Born et al. 1986, Gronfier et al. 1999) but not all (V gontzas et al. 1997) recent studies of endogenous cortisol-sleep rhythms, lower cortisol secretion rates were associated with increased sleep bouts. The findings of these studies, in conjunction with ours, further support the view that low minimal levels of HPA/CRH activity foster better sleep.

Overall, these results indicate that higher 24-h UFC under baseline conditions in senior women is associated with earlier risetimes and less REM sleep; associations with NREM sleep were nonsignificant. This association with earlier risetime persisted when habitual risetime data (PSQI) were used rather than sleep lab risetimes. It could be hypothesized that the REM findings were secondary to the earlier risetimes in the laboratory setting. To examine this possibility, we conducted efficiency), in addition to REM and risetime. In multiple regressions controlling for risetime, REM-UFC correlations remained significant; however, the NREM sleep quality measures were no longer significant, indicating that these associations were secondary to the earlier risetime in the stress condition. These findings are consistent with the hypothesis that higher 24-h baseline secretion of endogenous free cortisol is associated with earlier risetimes and independently with reduced REM sleep, and that SWS-cortisol associations may be secondary to earlier risetimes in optimally healthy seniors.

Aging effects on stress reactivity are well established, particularly the more prolonged response following stressful events (Wilkinson et al. 1997). The present study revealed both sleep and cortisol responses to a mild stressor that were more pronounced in the non Estrogen Replacement Therapy group. Sleep was significantly decreased (confirming the findings of Moe et al. in press) and UFC significantly increased in response to the 24-h stress in the non Estrogen Replacement Therapy group; with lesser increased effects in the Estrogen Replacement Therapy group; further significant UFC-sleep correlations were absent in the Estrogen Replacement Therapy group. Together these findings indicate that Estrogen Replacement Therapy may ameliorate sleep and cortisol responses to mild stressors. Limited data on women on progesterone-opposed Estrogen Replacement Therapy indicated that progesterone may diminish some of the sleep protective effects of ERT.

It is also possible that sleep loss may lead to an enhanced stress response involving increased cortisol production. In a sleep curtailment study in young men, Leproult and colleagues (Leproult et al. 1997) observed higher cortisol levels on the evening following partial sleep curtailment. This is unlikely to explain the current findings, because the stress reactive non Estrogen Replacement Therapy group had similar or better baseline sleep than the nonstress reactive Estrogen Replacement Therapy group.

Earlier risetimes and altered REM sleep are known to occur with advanced circadian phase (Van Cauter et al. 1996), a phenomenon that is commonly observed in older individuals. The present findings indicate that stress reactivity, sleep impairment, and circadian phase advances may interact in older women, particularly those not on ERT. Recent work has shown that the circadian pacemaker in the SCN is connected anatomically and functionally with the paraventricular neu-rons controlling CRH and ACTH release (Gomez et al. 1997, Vrang et al. 1995), which are known to be stress activated. Together with the present findings, this indicates that stress reactivity may be an important factor contributing to age

related sleep and circadian impairments. Factors that influence stress reactivity may provide fruitful areas of study for future sleep and aging research.