|Home | About | Journals | Submit | Contact Us | Français|
Older adults have reduced sleep quality compared to younger adults when sleeping at habitual times, and greater sleep disruption when their sleep is at adverse times. The purpose of this analysis was to investigate how subjective measures of sleep relate to objectively-recorded sleep in older subjects scheduled to sleep at all times of day. We analyzed data from 24 healthy older (55–74 years) subjects who took part in a 32-day inpatient study where polysomnography (PSG) was recorded each night and subjective sleep was assessed after each scheduled wake time. The study included baseline nights and a forced desynchrony (FD) protocol when the subjects lived on a 20-hr rest-activity schedule. Our post-sleep questionnaire both included quantitative and qualitative questions about the prior sleep. Under baseline and FD conditions, objective and subjective sleep latency were correlated, subjective sleep duration was related to slow wave sleep and wake after sleep onset, subjective sleep quality was related to Stage 1 and 2 sleep, and sleepiness and refreshment at wake time were related to duration of premature awakening. During FD, most measures of objective and subjective sleep varied with circadian phase, and many additional correlations between objective and subjective sleep were present. Our findings show that when sleeping at habitual times, these healthy older subjects did not perceive their generally poor sleep quality, but under FD conditions where sleep quality changed from day-to-day their subjective sleep ratings were more associated with their objective sleep.
Changes in the timing, consolidation and maintenance of sleep have been observed to be associated with aging, even in healthy people without sleep disorders (Bliwise 1993; Dijk et al. 2001; Duffy et al. 1998; Ohayon et al. 2004; Reynolds III et al. 1985). Such changes in sleep quality with aging include reductions in total sleep time and amount of slow wave sleep, and increases in the amount of wakefulness during sleep, with more awakenings, particularly in the latter part of the night (Bliwise 1993; Dijk et al. 2001; Duffy et al. 1998; Ohayon et al. 2004; Reynolds III et al. 1985). The self-assessment of sleep quality by older subjects often differs from their actual sleep structure (Åkerstedt et al. 2002; Buysse et al. 1991; Suzuki et al. 2004; Vitiello et al. 2004). Such discrepancies between subjective sleep quality and objective assessments of sleep are manifested as either a poorer perception of sleep compared with the objective measurement, or a good perception of sleep when there is objective sleep disruption (Edinger et al. 2000). In the latter case, an individual who might need treatment does not seek it because they perceive their sleep as good,.
Objective sleep quality consists not only of the total duration of sleep, but also of the architecture of sleep (amount of the different sleep stages across the sleep episode), the amount of wake during the sleep episode, and the frequency and duration of awakenings across the night. In young adults, there is deep slow wave sleep (SWS, Stages 3 and 4) in the first few NREM-REM cycles, little wake during the night, and few awakenings once sleep has been initiated. In older adults there are changes in objective sleep, with decreased amounts of SWS, more wakefulness after sleep onset (WASO), and more transitions between sleep and wakefulness (Dijk et al. 1999; Ehlers and Kupfer 1997; Klerman et al. 2004), resulting in lower sleep efficiency. In addition to these age-related changes in the structure and consolidation of sleep, there are age-related changes in sleep timing, with bedtimes and waketimes advancing to an earlier hour, and the sleep of older adults is more disrupted when scheduled at adverse times of day, with this disruption primarily affecting the latter part of the sleep episode (Bliwise 1993; Dijk et al. 1999). Although sleep disorders in older people are often comorbid conditions of other age-related diseases (McCurry and Ancoli-Israel 2003;Ribeiro et al. 2002), otherwise healthy older individuals also show reductions in SWS and sleep efficiency and increases in awakenings (Carrier et al. 1997; Carrier et al. 2005; Ehlers and Kupfer 1997; Haimov and Lavie 1997; Ohayon et al. 2004). Age-related changes in sleep structure are seen even in middle-aged adults; stage 2 sleep is increased and SWS and sleep efficiency (SE) are decreased in comparison with younger adults (Carrier et al. 1997; Carrier et al. 2005; Ehlers and Kupfer 1997; Haimov and Lavie 1997; Ohayon et al. 2004).
Subjective reports of sleep quality are important in the clinical setting, and can help determine whether further screening and/or treatment for a sleep complaint might be warranted (Åkerstedt et al. 2002; Suzuki et al. 2004; Vitiello et al. 2004). Several sleep-rating measures have been developed to aid clinicians, and these questionnaires focus on subjective estimates of sleep duration, latency, waking during the night, and other factors that could impact sleep quality and duration, such as comorbid conditions and medications. One of the most widely used of such questionnaires is the Pittsburgh Sleep Quality Index [PSQI (Buysse et al. 1989)]. While useful as a general assessment of overall sleep quality, it was not designed to rate sleep quality on a particular night. In a study of PSQIs given to healthy younger and older subjects, the authors reported that the older subjects showed worse global PSQI scores, and the PSQI component scores for subjective sleep quality, sleep duration, sleep efficiency, and sleep disturbances were worse in the older than the young subjects (Buysse et al. 1991). Despite the fact that objective sleep quality was worse among the older subjects, a majority of them showed a global PSQI score in the good sleeper range. When a direct comparison between PSQI score and objective PSG data was made, there were no significant correlations between global or component scores on the PSQI and any objective sleep measure among the older subjects.
The Karolinska Sleep Diary (KSD) is another questionnaire that was developed to assess subjective sleep quality (Åkerstedt et al. 1994b). Comparisons between KSD scores and objectively-recorded sleep in young adults studied on multiple nights with varying sleep quality have been performed. Those analyses showed that overall subjective sleep quality was related more to sleep efficiency and continuity, but not to individual sleep stages, and that sleep efficiency in the young adults studied had to be >87% to be subjectively rated as good (Åkerstedt et al. 1994a). In addition, the authors reported that the ease of awakening was related to poor objective sleep quality. A factor analysis of KSD questions was conducted, and that analysis found that KSD questions fell into two general factors, with a Sleep Quality Index (SQI) related to a number of objective sleep measures, including slow-wave sleep, sleep efficiency, and total sleep time (Keklund and Åkerstedt 1997). While the KSD has been used in studies of sleep in young and middle-aged adults, it has not been examined in older adults.
While each of the subjective indices described above has clinical relevance, they were designed more to assess the large changes in sleep quality due to sleep or other medical disorders, rather than assessing the ‘normal’ sleep changes that may occur gradually with healthy aging. Furthermore, most of the sleep quality questionnaires are for rating overall sleep quality, not sleep quality on a single night. Because of this, and because of prior reports that there is greater discrepancy between subjective and objective ratings of sleep in older adults, we conducted the present analysis to examine the relationship between objective and subjective sleep in healthy older subjects. We chose to examine this in a study in which sleep had been recorded in the same individuals nightly for a month, and where those same subjects had completed a post-sleep questionnaire containing both quantitative and qualitative questions about the prior night’s sleep each morning. Our study protocol therefore allowed for multiple assessments within an individual subject, and was designed to produce a wide range of sleep quality.
To conduct our comparison of objective and subjective sleep in older adults, we chose a protocol in which healthy older participants lived in the laboratory for many days, had their sleep polysomnographically recorded each night, and completed our standard laboratory post-sleep questionnaire after scheduled awakening each morning. The study was designed to test the effects of pre-sleep melatonin administration on objective sleep. Results of our detailed analysis of melatonin on sleep stages and EEG spectra will be published elsewhere. For the purpose of the present analysis, all sleep episodes were considered without respect to the treatment condition, as we were mainly interested in the comparison of objective and subjective sleep measures across a range of sleep quality.
Subjects were 11 men (66.9 ± 6.7 years (mean ± standard deviation), range = 56 to 78) and 13 women (61.9 ± 5.1 years, range = 55 to 72) recruited from the community for a month-long inpatient study of the effects of exogenous melatonin administration on sleep. Subjects were healthy and not taking medications, and were free of medical, psychological, and sleep disorders. This was determined by medical history, physical examination by a physician, electrocardiogram, blood and urine chemistries; by the Minnesota Muntiphasic Personality Inventory (Webb and Friel 1971), Geriatric Depression Scale (Yesavage et al. 1983), and an interview with a clinical psychologist; and by an all-night clinical PSG. The study procedures were reviewed and approved by the Partners Health Care Human Subjects Committee, and were conducted in accord with the Declaration of Helsinki. Each subject gave written, informed consent prior to starting the study procedures.
The inpatient study duration was 32 days (see Figure 1). For the first 3 days (baseline), the subject was scheduled to sleep for 8 hours at their habitual times. This was followed by a “forced desynchrony” protocol where the subject was scheduled to live on a 20-hour rest-activity schedule consisting of 6.7 hours scheduled bed rest and 13.3 waking hours, for 30 cycles. Following this, the subject was scheduled for 3 24-hour days at the same clock hours as the baseline. Each study day, the subject followed a pre-set schedule, beginning with being verbally awakened at the scheduled time, taking a battery of performance tests for the first 45 minutes of waking (Silva and Duffy 2008), a brief shower after getting out of bed, three meals (breakfast, lunch, dinner), no lying down or napping, and remaining in bed in the dark throughout each scheduled rest episode.
Throughout the study, the subject lived in a private study suite without windows and did not have access to timing devices (clocks, watches) or other devices that could be used to tell the time of day (television, radio, mobile telephone, etc.). Staff members were trained to avoid discussion of time or time of day with the subject. Room temperature was maintained at ~24°C, the light level during scheduled wake episodes was ~0.0087 W/m2 (3.3 lux) at 137 cm from the floor in a horizontal direction, and was less than 0.048 W/m2 (15 lux) at 187 cm from the floor in a vertical direction. Core body temperature was collected at 1-minute intervals throughout the study using a rectal thermistor (Measurement Specialties, Inc., Hampton VA) in order to assess circadian period and phase. During the scheduled wake episodes, the subject was given a short performance testing battery approximately every 2 hours; results from those tests will be reported elsewhere (Silva et al. 2008). Other than the scheduled shower, meals, and performance tests, during the wake episodes the subject was free to pursue sedentary activities within the study room. This typically consisted of reading, watching videos, listening to music, or pursuing hobbies.
Thirty minutes before scheduled lights off, the subject received a pill which on all baseline nights and on 18 of the forced desynchrony sleep episodes contained placebo, and on the other 12 forced desynchrony sleep episodes (either the first 12 or final 12) contained melatonin (either 0.3 mg every night or 5.0 mg every night). The melatonin timing and dosage during the forced desynchrony segment was assigned in a randomized, double-blind, and gender-balanced manner. For the purpose of the present analysis, all forced desynchrony sleep episodes were considered without respect to the treatment condition, as we were mainly interested in the comparison of PSG and PSQ scores across a range of sleep quality.
For all scheduled sleep episodes, the polysomnogram (PSG) was recorded using a standard montage: electroencephalogram (EEG) recorded from four derivations (C3, C4, O1, O2) each referenced to the contralateral mastoid; electro-oculogram (EOG; LOC and ROC referenced to the contralateral mastoid); and submental electromyogram (EMG). PSG data were acquired using a Vitaport Digital Sleep Recorder (Temec Instruments, Kerkrade, Netherlands). About 2 hours prior to each sleep episode, a technician placed the electrodes. Approximately 1 hour before scheduled lights off, the subject got into bed in a sitting position and the PSG recording was started. The subject remained in bed in a sitting position until a few minutes before scheduled lights off when the bed was laid flat for sleep. During this pre-sleep time, the subject took 3 short test batteries and was attended by a staff member to ensure they remained awake until scheduled lights off. Staff members did not enter the subject’s room during scheduled sleep unless called by the subject (for a bedpan or urinal, for example). Thus, if a sensor became loose or an equipment problem occurred within the room, the subject was not disturbed by the staff and that signal(s) was lost.
After the study was complete, the PSG records were scored in 30 second epochs by a trained technician using standard methods (Rechtschaffen and Kales 1968). In order to be included in our analysis, at least 95% of the epochs between scheduled lights off and lights on had to be scorable (see below).
A specialized software program [TASCI File Manager] was used to organize the scored PSG data and to compile information from each sleep episode. Objective sleep data derived from the PSG records included sleep onset latency (SOL, defined as the first three consecutive epochs of sleep), total sleep time, minutes of each sleep stage (Stages 1–4 and REM), minutes of slow wave sleep (SWS, Stages 3 + 4), minutes of NREM sleep (Stages 1–4), minutes of wake, wake after sleep onset (WASO), minutes of movement time and undefined epochs, sleep efficiency, final epoch of sleep, and number and duration of awakenings.
Following each scheduled wake-up time, the subject was given a post-sleep questionnaire (PSQ) to complete before getting out of bed. This included four quantitative questions about the previous sleep episode [self-assessed sleep latency, sleep duration, time awake prior to scheduled rising time (each scored in minutes), and number of awakenings]. In addition, there were two questions related to subjective sleep quality rated on a 7-point scale (soundness of sleep, 1=extremely light to 7=extremely sound; overall evaluation of sleep, 1=extremely bad to 7=extremely good) and two questions related to current subjective feeling rated on a 7-point scale (current level of sleepiness, 1=not sleepy at all to 7=extremely sleepy; current feeling of refreshed, 1=not refreshed at all to 7=extremely refreshed). This questionnaire was developed in our laboratory more than a decade ago and has been routinely administered to participants in a wide rnage of studies. We reported previously that responses on one of the PSQ questions vary with circadian phase (Duffy et al. 1998) and that this circadian pattern of response mirrors that of objective sleep (Dijk and Duffy 1999). However, no night-by-night by comparison of the PSQ with PSG has been conducted previously.
We first examined the questions on the PSQ in order to determine whether the subjects’ responses on each question were independent. To do this, we pooled all the nights for which we had PSQs and then performed a Spearman correlation analysis comparing responses on each PSQ question to responses on all the other questions.
We calculated mean (± standard deviation) values for each objective and subjective sleep variable separately for the baseline and forced desynchrony (FD) conditions. These baseline and FD means were compared using a Student t-test. Because there were 7 subjective sleep responses and 13 objective sleep measures, we conducted these t-tests tests at a Bonferroni-corrected level (α=0.05/20=0.0025). In order to determine what aspects of objective sleep were related to subjective sleep, we next performed a correlation analysis. We did this for the baseline and FD nights separately using a Spearman correlation coefficient.
To explore whether there was a gender difference in objective sleep quality or in the way the subjects rated their sleep, we compared baseline objective and subjective sleep variables between the men and women using a Student t-test, and as above conducted them at a Bonferroni-corrected level (α=0.05/20=0.0025)..
Finally, we conducted an analysis to determine whether the subjective responses on the PSQ varied with circadian phase. To do this, we first assessed the period of the core body temperature data from the FD portion of the study using non-orthogonal spectral analysis [NOSA; (Czeisler et al. 1999)], and using that period information and the phase of the overall temperature rhythm on the first FD day (designated circadian phase 0), we assigned a circadian phase to each scheduled wake time (and its associated PSQ and prior sleep episode). We binned the data into eight 45-degree circadian bins, with each bin representing three circadian hours. Variation with circadian phase for each of the objective sleep measures and the subjective responses on the PSQ were then assessed using a mixed model, which quantifies both between-subjects (i.e., inter-individual) variability and within-subjects (i.e., temporal) variability (Laird and Ware 1982). We used an ar(1) correlation structure to account for the within-subject variance. Because there were 7 subjective sleep responses and 13 objective sleep measures, we conducted these tests at a Bonferroni-corrected level (α=0.05/20=0.0025).
All analyses were performed using SAS v 6.12 (SAS Institute, Cary NC).
In our analysis of how responses on the PSQ questions were related, we found a high correlation between responses to the ‘soundness of sleep’ and ‘evaluation of sleep’ questions (r = 0.88, p<0.0001). From this, we concluded that these two questions were not independent, and for our subsequent analyses we only included responses to the ‘evaluation of sleep’ question.
There were 72 nights of baseline sleep from the 24 subjects. Data from all subjects for Night 1 were omitted from our analysis due to the so-called ‘first night effect’ (Agnew, Jr. et al. 1966). Data from the PSQs on Nights 2 and 3 were not available for one subject, and PSG data from Night 2 for two subjects were omitted from analysis due to more than 5% of the epochs being unscorable because of equipment or sensor problems. We therefore included a total of 44 baseline sleep episodes from 23 of the subjects in our analysis.
The objective baseline sleep was similar in composition to what has been reported for healthy older adults, and was comprised of 60.6 ± 6.3% NREM, 18.1 ± 5.2% REM, and 20.9 ± 8% wakefulness (see Table 1). Subjects evaluated their baseline sleep overall as between ‘average’ and ‘good’ (4.73 ± 1.25), and stated that they felt between ‘somewhat refreshed’ and ‘refreshed’ (3.64 ± 1.53) and only ‘slightly sleepy’ (2.02 ± 0.96) at wake time.
When the objective and subjective assessments of baseline sleep were assessed, there was a significant correlation between subjective sleep onset latency (SOLs) and objective sleep onset latency (SOLo) (see Table 2). Subjective evaluation of sleep was negatively correlated with the duration of Stage 1 sleep and positively correlated with the duration of Stage 2 sleep (see Figure 2 and Table 2). Total perceived sleep time was correlated with the duration of slow-wave sleep and negatively correlated with Stage 1 and WASO (see Table 2). Decreased subjective sleepiness and increased refreshment after wake time were associated with the amount of wakefulness prior to scheduled wake time (see Table 2).
Because of our stringent criteria (α=0.05/20=0.0025), when we compared the male and female subjects on the baseline nights we did not find significant gender differences in any objective or subjective sleep measure. We did observe a trend for the female subjects to have more SWS than the males (61.4 ± 33.21 vs. 39.68 ± 33.03 minutes, p = 0.036). The females also tended to have more objective awakenings (32.58 ± 18.23 vs. 22.65 ± 9.11, p = 0.025) and tended to have shorter duration awakenings (3.44 ± 2.53 vs. 5.10 ± 2.99 minutes, p = 0.052). The females tended to report fewer subjective awakenings than the males (1.79 ± 1.18 vs. 2.68 ± 1.0, p = 0.011).
There were 30 scheduled sleep episodes during the FD condition per subject. One subject did not complete the final FD sleep episode because she became ill and her study was ended early, and the PSG file from one sleep episode from another subject was lost. A total of 13 sleep episodes from 8 of the subjects were not included in our FD analysis due to having more than 5% of the epochs unscorable, and 3 sleep episodes from 2 other subjects were not included because of problems with the PSQ. This resulted in a total of 702 FD sleep episodes available to be included in our analysis, with each subject contributing between 28 and 30 FD sleep episodes to the analysis.
Objective sleep quality varied widely throughout the FD condition, with sleep efficiency averaging more than 6% lower (p < 0.0001) than in the baseline condition (see Table 1). Objective sleep latency was not significantly different between the baseline and FD conditions, nor was the duration or percentage of SWS. There were fewer minutes of Stage 2 and NREM sleep in the FD than in the baseline condition (see Table 1), and the percentage of these sleep stages was also less in the FD condition than in the baseline (Stage 2: 34.6 ± 10.9% vs. 38.8 ± 8.4%, p < 0.01; NREM: 55.5 ± 12.4% vs. 60.6 ± 6.3 %, p < 0.0001). There were also fewer minutes of Stage 1 and REM sleep in FD than in baseline (see Table 1), but the overall percentage of these sleep stages was not significantly different between FD and baseline (Stage 1: 9.5 ± 5.7% vs. 11.0 ± 5.9%, p = 0.08; REM: 17.1 ± 6.7% vs. 18.1 ± 5.2%, p = 0.26). Overall, during the FD condition the subjects spent significantly longer lying awake just prior to scheduled wake time than they did in the baseline condition (p < 0.0001; see Table 1). While the overall duration of wake during the scheduled sleep episode was not significantly different in the FD and baseline conditions, it was a significantly greater percentage of the scheduled sleep time during the FD condition (27.2 ± 16.5% vs. 20.9 ± 8.0%, p < 0.0001).
Subjects rated their subjective sleep latency, total sleep time, and number of awakenings, overall evaluation of sleep, and feeling of refreshment at wake time similarly in both FD and baseline conditions (see Table 1). Subjects estimated their early awakenings as being longer during FD than during baseline (p < 0.01), and rated themselves as sleepier at wake time in the FD condition (p < 0.01), although these latter two subjective measures did not reach our stringent criteria for significance.
The correlation between objective sleep and the subjective assessment of sleep under FD conditions was performed in the same manner as for the baseline nights. Those significant associations between objective and subjective sleep in the baseline condition were also significant in the FD condition, although in most cases there was a weaker association between the factors (see Table 2). These included the correlation between subjective and objective sleep onset latency; total perceived sleep time and Stage 1, SWS, and WASO; subjective evaluation of sleep and duration of Stage 1 and Stage 2; sleepiness at wake time and SOLo; and refreshment at wake time and amount of wakefulness prior to scheduled wake time (see Table 2).
In addition, there were associations between subjective and objective sleep that were not significantly correlated at baseline, but which became significantly correlated during the FD condition. Subjective total sleep time was significantly correlated with objective TST, with duration of NREM sleep and with sleep efficiency (see Figure 3), and was negatively correlated with scored wakefulness (see Table 2). Subjective estimation of the duration of wake prior to scheduled wake time was positively correlated with objectively-scored wakefulness prior to scheduled wake time (see Figure 4), and WASO. Subjective wake prior to scheduled wake time was also associated with objective TST, sleep efficiency and duration of Stage 2 (see Table 2).
When we examined the influence of circadian phase on all-night objective sleep parameters, we found a significant influence of circadian phase on sleep latency (F7, 671 = 8.68, p < 0.0001), sleep efficiency (F7, 671 = 24.73, p < 0.0001), NREM (F7, 671 = 15.69, p < 0.0001), and REM sleep (F7, 671 = 24.25, p < 0.0001), as well as in the amount of wakefulness at the end of the night (F7, 671 = 11.41, p < 0.0001), the total amount of wakefulness during the scheduled sleep episode F7, 671 = 26.33, p < 0.0001), and the average duration of awakenings (F7, 671 = 8.06, p < 0.0001). With the exception of subjective number of awakenings (F7, 671 = 0.66, p = 0.7), all subjective sleep estimates (latency: F7, 666 = 4.15, p < 0.001; total sleep time: F7, 671 = 3.90, p < 0.001; duration of waking prior to scheduled time: F7, 669 = 13.40, p < 0.0001; overall evaluation of sleep: F7, 671 = 5.35, p < 0.0001; feeling of sleepiness at wake time: F7, 671 = 10.85, p < 0.0001; feeling refreshed at wake time: F7, 670 = 3.73, p < 0.001) also showed significant circadian variations.
Because of these significant circadian variations in objective and subjective sleep during the FD, we conducted a second comparison of FD vs. baseline sleep and restricted FD sleep episodes to those which occurred at similar circadian phases to the baseline nights. The phase at wake time during baseline ranged from 10.75 to 92.25 degrees between the subjects, and there were 152 FD sleep episodes in this range. When we re-did our comparison of baseline vs. FD sleep on this restricted data set, we found that objective sleep latency, and the percentage of Stage 1, Stage 2, SWS, and NREM sleep were similar in the two conditions. The percentage of REM sleep was significantly higher in the FD condition (22.08 ± 5.0% vs. 18.06 ± 5.2%, p < 0.0001) as was sleep efficiency (84.15 ± 8.22% vs. 79.05 ± 8.03%, p < 0.001). While the number (p = 0.043) and average duration (p = 0.011) of awakenings did not reach the stringent significance level we set of < 0.0025, they were lower in the FD condition, resulting in the overall percentage of wake being significantly lower in the FD condition than in baseline (15.81 ± 8.2% vs. 20.89 ± 8.02%, p < 0.001). Subjective ratings of total sleep time and number of awakenings did not differ between the two conditions, nor did the subjective feeling of refreshment at wake time. Subjective sleep latency (11.22 ± 7.91 min. vs. 19.22 ± 20.71 min. p = 0.017) and subjective duration of premature awakening (10.39 ± 21.67 min. vs. 19.43 ± 29.57 min., p = 0.064) tended to be shorter in the FD condition, and subjects rated themselves as feeling slightly more sleepy at wake time (p < 0.001) and evaluated their sleep as significantly better in the FD condition (p < 0.01).
The purpose of this study was to investigate the relationship between subjective assessments of sleep and objectively-recorded sleep in older subjects using multiple assessments in each participant, and with sleep scheduled at many times of day in order to produce a range of objective sleep quality. Our study protocol allowed us to examine the relationship between objectively recorded sleep and subjective estimates of sleep in 24 healthy older adults on two baseline nights scheduled at each subject’s habitual times, and on 28–30 subsequent nights when sleep was scheduled across circadian phases. We found that several aspects of subjective sleep estimates were significantly associated with objective sleep in both baseline and FD conditions. Among the quantitative measures we assessed, we found that subjective and objective sleep latency were correlated, and that longer estimations of subjective TST were associated with greater amounts of SWS and lesser amounts of Stage 1 and WASO in both baseline and across the FD condition.
When more qualitative measures of subjective sleep were examined, we found that overall evaluation of sleep quality was positively related to duration of Stage 2 and negatively related to duration of Stage 1. When subjects were asked about their feelings of sleepiness and refreshment at wake time, we found that longer durations of lying awake before scheduled wake time led to greater refreshment and less sleepiness in both conditions.
Our finding of an association between subjective and objective sleep latency is consistent with prior reports from both young and older adults (Åkerstedt et al. 1997; Hoch et al. 1987), as is our finding that subjective TST was associated with duration of SWS (Åkerstedt et al. 1997; Hoch et al. 1987; Keklund and Åkerstedt 1997). The association between subjective TST and duration of SWS was significant although not strong, and occurred in both conditions despite SWS representing only about 10% of the scheduled sleep time.
We found that how sleepy or refreshed the subjects felt upon awakening was associated with how long they had been lying awake before their scheduled wake time. Longer durations of prior wakefulness were associated with less sleepiness and a greater feeling of refreshment at wake time. This finding is consistent with a prior report from young subjects scheduled to sleep at many different times of day (Åkerstedt et al. 1994a), as in our study. It may be that subjects felt more refreshed and less sleepy because lying awake in bed allowed any sleep inertia they might have experienced to be dissipated by the time scheduled wake time (and their assessment of sleepiness/refreshment) occurred. This relationship between lying awake in bed and feeling less sleepy and more refreshed may also have been due to the unforced nature of these awakenings.
Additional associations between objective sleep and subjective ratings of sleep were identified during the FD condition when compared with the baseline nights. These included both quantitative and qualitative measures of sleep. This is likely due to the greater range of objective sleep quality in the FD condition than in the baseline. In particular, we found significant associations between subjective estimates of waking prior to scheduled wake time and several objective sleep measures, between subjective and objective number of awakenings, and between subjective estimates of TST and objective sleep measures. When more qualitative aspects of subjective sleep were examined, we also found additional associations with objective sleep during the FD.
Many of the associations between objective sleep and subjective assessment of sleep that were significant in the FD condition were related to the subjective estimation of the duration they were lying awake prior to scheduled wake time. In a previous study, we analyzed the overall circadian rhythm of this subjective assessment in young and older adults (Duffy et al. 1998). While in that study we did not compare night-by-night subjective and objective assessments, we found that the overall circadian rhythm paralleled that of objective wakefulness within scheduled sleep in both age groups (Dijk and Duffy 1999). Given our previous finding that the sleep of even healthy older adults is markedly disrupted when it occurs at an adverse circadian phase, and that this disruption is strongest towards the latter part of the sleep episode (Dijk et al. 1999), our current finding of a strong night-by-night correlation between subjective premature awakening duration and objective sleep parameters is not surprising.
Our subjects reported their baseline sleep as being of good quality even though their sleep efficiencies and amount of SWS were rather low compared with sleep in healthy young adults. This finding is consistent with prior reports from healthy older subjects (Buysse et al. 1991; Vitiello et al. 2004). Buysse (1991) has argued that the gradual changes in sleep quality associated with aging may lead the older individual to adapt their perception of sleep quality to the actual changes in their sleep structure, and thus not recognize that their typical sleep is actually disrupted [see also (Vitiello et al. 2004)]. The conditions of FD, whereby abrupt changes in sleep quality occur from night to night, likely allowed our subjects to better perceive the relative changes in their sleep that allowed associations between overall evaluation of sleep and sleep efficiency, amount of wakefulness during scheduled sleep, and TST to be made.
Our finding that there are few and only modest associations between subjective ratings of sleep and objective sleep measures on baseline nights is also consistent with our recent report on objective sleep quality in a larger group of healthy young and older adults (Pavlova et al. 2008). In that study, we conducted an overnight clinical polysomnographic exam on healthy non-complaining subjects and found that more than half of those over age 50 had evidence of clinically-significant sleep-disordered breathing. It is likely that these non-complaining older subjects with sleep-disordered breathing had experienced gradual changes in their sleep quality over several years, and had therefore adjusted their perception of normal sleep quality over this time, as has been suggested occurs with typical age-related reductions in sleep quality (Buysse et al. 1991).
Previous studies of sleep at night in older adults have found that the sleep of older women is of better quality than that of older men, although older women have been reported to have more complaints about their sleep than older men (Vitiello et al. 2004). In our study, the only gender difference in baseline objective sleep was a tendency for the older men to have less SWS and fewer awakenings. In the subjective sleep assessments on the baseline nights, the older men actually reported more awakenings than the women, with no other gender differences.
This group of healthy older adults had baseline sleep typical for their age group, and some aspects of their quantitative subjective sleep ratings were associated with aspects of their PSG-staged sleep. There were also modest associations between their self-ratings of sleep quality and the amount of the lighter NREM stages of sleep. The overall evaluation of their baseline sleep was on average ‘good’, despite the generally low baseline sleep efficiencies. Our findings are consistent with prior reports from older adults sleeping at habitual times, and also consistent with data obtained from younger adults scheduled to sleep at all different times of day. Our analyses, together with prior reports, demonstrate that subjective sleep quality is not easily quantified in this age group, and thus simple questionnaires are unlikely to be able to identify normal age-related reductions in sleep quality that occur gradually over many years, nor gradual changes in sleep quality due to moderate sleep-disordered breathing (Åkerstedt et al. 1994b; Buysse et al. 1989; Harvey et al. 2008).
There are several limitations to our study findings. First, our subjects were highly screened to be extremely healthy and to not have clinically-significant sleep-disordered breathing. As such, they are not representative of their age group, many of whom are on medications, have chronic medical conditions, and/or have undiagnosed sleep disorders. In addition, while we did have multiple assessments per subject, our overall study population was small, and this may also limit the applicability of our findings. Next, subjects in our study were required to remain in bed in the dark until their scheduled wake time, and were given the PSQ to complete after that scheduled wake time. Therefore, they did not complete the PSQ a consistent amount of time after final awakening, and this may have affected their ability to rate their sleep, and this in turn could have affected the association between their sleep ratings and the PSG measures. Our study subjects were also scheduled to sleep and wake across circadian phases during the FD portion of their study. While this allowed us to assess the subjective-objective sleep ratings across a wide range of sleep quality, most people sleep at a much more restricted range of phases, and thus our Baseline findings are more relevant to “real world” situations than are our FD findings. Furthermore, during the FD subjects were also awakened at times they would not normally wake at under entrained conditions; this may also have impacted their ability to remember their previous night of sleep. Finally, subjects in our study did not sleep well on many of the FD nights because of the time at which they were scheduled to sleep. That sleep restriction may have influenced their subsequent PSQ responses, by affecting their memory of their sleep.
The authors wish to thank the study participants; D. McCarthy and C. O’Brien for subject recruitment; M.J. Duverne-Joseph, R. Webb, V. Hobbes, and V. Sparkes of the Division of Sleep Medicine (DSM) Chronobiology Core for assisting with the data collection and processing; the Brigham and Women’s Hospital General Clinical Research Center (BWH GCRC) staff; B.J. Lockyer, E. Riel, and G. Renchkovsky of the DSM Sleep Core for scoring the sleep recordings; and C.A. Czeisler for overall support. Supported by grants P01 AG09975 (to C.A. Czeisler) and M01 RR02635 (BWH GCRC) from the US National Institutes of Health. MM was supported by fellowships from the Novartis and La-Roche Foundations, Switzerland; DO’D was supported by a FAS Science Challenge Internship funded by the Irish government.
The following authors report no actual or potential conflict of interest: Deirdre O’Donnell, Edward J. Silva, Mirjam Munch, Joseph M. Ronda, Wei Wang, Jeanne F. Duffy