- Email: [email protected]
Slow-wave sleep and waking cognitive performance II
Physiology & Behavior 70 (2000) 127–134
Slow-wave sleep and waking cognitive performance II: Findings among middle-aged adults with and without insomnia complaints Jack D. Edingera,b,*, D. Michael Glennc,d, Lori A. Bastiana,b, Gail R. Marshb a
Veterans Affairs and bDuke University Medical Centers, Durham, NC, USA c John Umstead Hospital, Butner, NC, USA d University of North Carolina, Chapel Hill, NC, USA Received 9 June 1999; received in revised form 8 November 1999; accepted 10 February 2000
Abstract Previous studies showing a relationship between nocturnal slow-wave sleep (SWS) and subsequent diurnal performance among young normal sleepers and older insomnia sufferers have provided limited support for the notion that this sleep stage serves a restorative role for neurocognitive functioning. The current study, which examined the relationship between SWS and reaction time performance among middle-aged adults with and without insomnia complaints, was conducted to further explore this possibility. A sample of 31 noncomplaining middle-aged (ages 40 to 59 years) normal sleepers and a like-aged sample of 27 insomnia sufferers, provided data for the current investigation. All participants underwent nocturnal sleep monitoring immediately prior to undergoing a battery of daytime tests that measured simple reaction time, vigilance/signal detection, and complex reaction time. Results showed relationships between reaction time performances on some tasks and some SWS measures among both the normal sleepers and insomnia sufferers. Findings supported our prediction that the presence of sleep pathology (e.g., insomnia) alters the SWS–performance relationship observed, but the results failed to show a consistent relationship between SWS and subsequent performance within either sample. The findings suggest that the specific performance demands of the task in question as well as physiological parameters other than SWS may determine performance as well. Findings for this and previous studies do provide some support for the contention that the neurocognitive restorative value of SWS may change across the lifespan. Possible implications of the study’s findings are discussed and directions for future research are considered. © 2000 Elsevier Science Inc. All rights reserved. Keywords: Slow-wave sleep; Waking cognitive performance; Middle-aged adults
1. Introduction Numerous studies [7,10,13,23–25] of sleep deprivation, growth hormone secretion, acute exercise, and physical fitness have implied that slow-wave or deep sleep may contribute to somatic restoration among various age groups. Although it has been theorized that nocturnal slow-wave sleep (SWS) might also serve a restorative role for neurocognitive functions, previous studies exploring the relationship between SWS and diurnal performance [3,8,11,14,17,21] have provided mixed results. Perhaps this variability is understandable inasmuch as the complexity of neurocognitive tasks employed in these studies has varied appreciably. Nonetheless, even those studies using similar performance tasks have produced inconsistent findings. Jurado et al. [14], * Corresponding author. VA Medical Center, Psychology Service (116B), 508 Fulton Street, Durham, NC 27705. Tel.: (919) 286-0411 ext. 17054; Fax: (919) 416-5832. E-mail address: [email protected].
for example, found a significant relationship between the amounts of SWS that young healthy adults obtained at night and their subsequent diurnal reaction times. In contrast, we [8] failed to replicate this finding among older (age ⭓60 years) adults with and without insomnia complaints. However, using slow-wave measures derived from computerassisted spectral analyses, we found that only insomnia sufferers showed an inverse relationship between daytime reaction time and EEG power in the 2 to 4-Hz spectral band. Hence, the potential restorative value of SWS for even fairly simple neurocognitive functions like reaction time remains questionable. Of course, one hypothesis suggested by these latter studies is that the degree to which slow-wave sleep restores neurocognitive processes among normal sleepers changes as a function of aging. Although Jurado et al. [14] found a strong relationship between nocturnal SWS and subsequent diurnal performance in healthy young normal sleepers, we and others failed to show a similar relation-
0031-9384/00/$ – see front matter © 2000 Elsevier Science Inc. All rights reserved. PII: S0031-9384(00)00 2 3 8 - 9
128
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
ship among senior adults. This observation as well as the fact that normal sleepers show marked declines in SWS across the lifespan [1,4] suggests that the degree to which SWS contributes to neurocognitive restoration may gradually decline through adulthood. Indeed, by late life, SWS may become what some investigators have called a “functionally meaningless remnant” [21]. Unfortunately there currently is no longitudinal data to corroborate this speculation. Furthermore, because previous studies of the SWS– neurocognitive performance relationship have generally ignored middle-aged normal sleepers, there currently is insufficient published cross-sectional data to support this contention. Thus, studies of the SWS–performance relationship among middle-aged normal sleepers would help fill this void in this literature. In addition, our previous findings suggest that conditions such as chronic primary insomnia (PI), may alter the observed relationship between SWS and neurocogntive functioning. As we and others [6,9,22] have noted, PI is typically perpetuated by aberrant sleep habits and conditioned arousal, which interfere with normal homeostatic regulation of the sleep process. As a consequence, those who suffer from PI experience enhanced night-to-night variability in SWS and other sleep parameters. Indeed, such individuals typically show sleep patterns characterized by one or more nights of poor sleep (i.e., partial deprivation), followed by one or more nights of good or extended sleep (i.e., recovery). Despite or, perhaps, due to their chronically unpredictable sleep, PI sufferers often appear hyperalert or neuocognitively aroused during the daytime [5,16]. The combination of erratic nocturnal sleep and diurnal hyperalertness/arousal could arguably alter the SWS–diurnal performance relationship that is observed among like-aged normal sleepers. Whereas our previous study with older adults suggests this may be the case, additional tests of this speculation seem warranted. The current investigation was conducted, in part, to explore the possible restorative role of nocturnal SWS for subsequent neurocognitive, reaction time functioning among middle-aged normal sleepers. In addition, this study was conducted to replicate our previous findings, which suggested that the nocturnal SWS–diurnal performance relationship observed in these normal sleepers varies from that observed within an age-matched sample of insomnia sufferers. To accomplish these study objectives, we extracted sleep and reaction time performance data obtained in a larger study of middle-aged adults with and without insomnia complaints. For the purposes of this study, we hypothesized that: (1) middle-aged normal sleepers would show an inverse relationship between the amount and intensity of their nocturnal SWS and their subsequent diurnal reaction times; (2) this SWS–reaction time relationship would be less pronounced among these middle-aged normal sleepers than that previously observed among young adult normal sleepers (e.g. [14]); and (3) middle-aged PI sufferers would show a more pronounced SWS–reaction
time performance relationship than would our age-matched normal sleepers.
2. Materials and methods 2.1. Subjects Data gathered for the current investigation were obtained from a cohort of middle-aged (40 to 59 years) adults while they participated in a larger study designed to address a number of research objectives in addition to those pertinent to the current report. Participants included in this study were drawn from samples of 32 (16 women, 16 men) middleaged normal sleepers and 32 age- and gender-matched insomnia sufferers who were recruited through announcements posted within two medical centers (i.e., Durham VA and Duke University Medical Centers) and via recruitment letters sent to Durham VA Medical Center volunteer workers and to individuals listed in the subject pool of the Duke Center for the Study of Aging and Human Development. The normal sleepers had no acute, sleep-disruptive medical disorders, reported no sleep complaints, and did not meet criteria for any sleep disorder on the basis of a Structured Interview for Sleep Disorders (SIS-D) [20]. The insomnia sufferers were also medically healthy individuals who: (a) had a history of insomnia complaints 艌6 months duration; (b) met SIS-D criteria for Primary Insomnia (PI); and (c) did not meet SIS-D criteria for any other sleep disorder. Excluded from the study were individuals who: (a) had a terminal illness; (b) had an unstable medical condition or a medical disorder (e.g., rheumatoid arthritis, thyroid disease, asthma) that compromises sleep; (c) had abnormal TSH levels on a screening thyroid panel (d) had a history of psychiatric illness(es); (e) met criteria for a major psychiatric (Axis I) condition on the basis of a Structured Clinical Interview for psychiatric disorders (SCID) (1); (f) were habitual substance abusers; (g) used sleep medications and were unwilling/unable to abstain from this practice; (h) were taking anxiolytics, antidepressants, or any other psychotropic medication; or (i) had 艌15 events per hour characterized by a total cessation of breathing (i.e., apneas) and/or dramatic reduction in airflow (i.e., hypopneas ⫽ 50% or greater decrease in airflow) on a screening night of sleep monitoring. Of all individuals enrolled in our research program, three participants (two male normal sleepers and one female insomnia sufferer) were dropped from consideration for the current investigation because their sleep recordings for the night before performance testing were of such poor technical quality that they were deemed unscoreable. We also noted that some performance data were lost from three male insomnia sufferers as a result of technician error in storing these data into computer memory. As a result, each of these individuals had missing performance data for some of the testing trials. Because our experience suggests that individuals tend to show both practice and circadian fluctuations in their responses to the performances tests used in this study,
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
we assumed that averaged performance data taken from individuals with missing data for some trials might not accurately represent their neurocognitive functioning across the day. Given this rationale, data from these three participants with missing data were eliminated from consideration in the analyses conducted in this investigation. Our remaining sample consisted of 31 (16 women, 15 men) normal sleepers (mean age ⫽ 46.4 years, SD ⫽ 5.2 years) and 27 (15 women, 12 men) PI sufferers (mean age ⫽ 49.7 years, SD ⫽ 5.7 years). As a group, those with insomnia had, on average, completed 15.3 (SD ⫽ 2.8 years) years of education and had had subjective sleep complaints for an average of 10.5 years (SD ⫽ 7.8 years). Three (5.2%) of these participants complained exclusively of sleep-onset difficulties, 12 (44.4%) had solely sleep-maintenance complaints, 11 (40.7%) reported a mixture of sleep onset and maintenance problems, and the remaining subject reported problems with chronically poor sleep quality. At the time of their study participation, 20 (74.1%) of the insomnia sufferers reported no current use of sedative hypnotics (e.g., benzodiazepines, zolpidem, etc.), 3 (11.1%) reported use of such agents one time per week or less often, 1 (3.7%) used such agents less than two times per week, and the remaining 3 (11.1%) used such agents two to four times per week to aid their sleep. The normal sleepers had comparable educational histories (mean years of education ⫽ 16.3 years; SD ⫽ 2.6 years) but reported no sleep difficulties and no current use of hypnotic sleep medications. For the purposes of this study, all participants using sedative hypnotics abstained from these medications during a 2-week wash-out period as well as during their participation in the study. 2.2. Polysomnographic sleep monitoring All sleep monitoring studies were conducted with Oxford Medilog 9000/9200 (Oxford Medical, Inc., Clearwater, FL) ambulatory recording devices. The Medilog is a cassette system with the capability of recording eight channels of electrophysiological data in analog format, as well as digital time in 1-s intervals, and subject-labeled events. Our monitoring montage for the current study included two EEG channels (C3-M2, Oz-Cz), one chin EMG channel, two channels to monitor eye movements, EOG (left eye—M1, right eye—M2), and two channels to monitor anterior tibialis EMG (right and left legs). In addition, respiration (nasal/ oral air flow–thermistor) was monitored to screen for apneic events. For purposes of the above-mentioned larger study, all individuals included in the current study underwent a total of 6 nights of sleep recording with either a series of three home (HPSG) or three lab (LPSG) recordings occurring initially in counterbalanced order. For all study participants, daytime performance testing occurred after the first series of three PSG studies (either HPSG or LPSG). By random assignment, one-half (16 normals, 16 insomniacs) of the 64 participants enrolled underwent their three HPSG studies prior to daytime testing, whereas the other half of the sample underwent three LPSGs prior to daytime testing. Of
129
these, the 58 who met selection criteria for the current study consisted of 30 (15 normal sleepers, 15 insomnia sufferers) who underwent HPSG prior to daytime testing and 28 (16 normal sleepers, 12 insomnia sufferers) who underwent LPSG prior to this testing. Once these individuals had completed all experimental procedures, sleep data from the night of monitoring immediately preceding daytime testing (i.e., third night) were extracted for the purposes of this study. Because previous studies [2,12] and our own experience [8] have suggested that computer-assisted EEG spectral analyses may provide a more sensitive characterization of slow-wave sleep than do conventional manual scoring techniques, we used this method in addition to conventional manual scoring techniques [18] to score our participants’ sleep recordings. For each recording, a Fourier transformation of the sleep data and subsequent spectral analysis was used to quantify the: (a) % of EEG power within the 0.5 to 2-Hz bandwidth during NREM sleep; and (b) % of EEG power within the 2 to 4-Hz bandwidth during NREM sleep. The computer software used was designed to exclude periods of wakefulness, stage 1, and REM sleep as well as movement artifacts from the two spectral measures derived. These spectral analytically derived measures along with estimates of SWS % (i.e., [SWS-min/time in bed-min] ⫻ 100%) derived from conventional scoring methods [18] served as the primary SWS measures used to test the study hypotheses. 2.3. Performance testing All study participants completed four 16-min trials composed of a series of three computer-administered reaction time tasks selected from the Neurobehavioral Evaluation System (NES) [15]. During the test administration, each participant was placed individually in a testing room, and was seated in front of a PC computer that contained the NES software used for the testing. At the beginning of each test, the computer software provided written instructions on the computer screen. The study participant read the instructions and then proceeded with the computer-guided standardized battery as outlined in the ensuing discussion. 2.3.1. Simple Reaction Time Test The first and least difficult test presented in each trial was the Simple Reaction Time Test (SRT). During this test, the participant was required to press a specially marked key on a computer keyboard whenever a figure (i.e., a small square) appeared on the computer screen. Once the test was begun, the figure appeared at intervals varying between 1000 and 2500 ms. Upon each presentation, the figure remained on the screen either until a response occurred or 1000 ms had elapsed. The total trial lasted approximately 5 min, and consisted of 90 presentations of the target stimulus. For each presentation of the figure, the computer software automatically recorded the time (in ms) between the appearance of the figure on the screen and the computer key-press response. For each 5-min SRT trial, the computer software
130
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
computed a mean response latency and a within-subject standard deviation of the participant’s response latencies. 2.3.2. Continuous Performance Test This somewhat more challenging test consisted of a signal detection task during which a target (i.e., the letter “S”) and background signals (i.e., the letters “A,” “C,” “E,” and “T”) were presented in rapid fashion on the computer screen. Once the test was begun, target and background letters appeared in a random sequence during which a 1:4 target-to-background ratio was maintained. Target and background letters were presented at the rate of 1 per second, and each letter remained on the computer screen for 50 ms. The total test lasted approximately 5 min, and included 60 presentations of the target and 240 presentations of the background stimuli. In responding to this test, the participant was required to press a specially marked key on the computer keyboard when and only when the preidentified target letter appeared on the screen. For each trial of this test, the NES software derived a mean response latency and a within-subject standard deviation for the response latencies. 2.3.3. Switching Attention Test The final test, the Switching Attention Test (SWAT), lasted approximately 6 min, and included three subtests. Each subtest required the participant to press specially marked keys on the right and left side of the keyboard in response to stimuli presented on the computer screen. During part I (side condition) of the SWAT, a square appeared either on the right or left side of the computer screen and the participant was required to press a marked key on the corresponding side of the computer keyboard. Upon each presentation, the stimulus remained on the screen either until a response occurred or 2500 ms had elapsed. This portion of the SWAT included 6 practice and 16 test presentations of the stimulus. During part II (direction condition), an arrow, pointing right or left, appeared in the center of the screen, and the participant was required to make right or left side key presses in response to the direction in which the arrow was pointing. As in the previous section of the SWAT, the stimulus remained on the screen either until a response occurred or 2500 ms had elapsed. This portion of the test included 4 practice and 16 test presentations of the stimulus. Finally, during parts IIIA and IIIB (switching conditions), an arrow (pointing right or left) appeared either on the right or left side of the screen. Preceding each presentation of this arrow by 1000 ms, one of two command words, “SIDE” or “DIRECTION,” appeared on the screen. This command word served to instruct the participant to respond by pressing a key on the side of the keyboard corresponding either to the side of the screen on which the arrow appeared or the direction in which the arrow was pointing. On 50% of the presentations the side of the screen on which the arrow appeared and the direction in which it was pointing agreed. On the remaining presentations, these two stimulus characteristics were in conflict. Throughout the test, these nonconflict and conflict presentations occurred in a random se-
quence. Overall, the switching condition included 8 practice and 48 test presentations of the command-stimulus combination. For each test trial, the NES computer software computed a mean response latency and a within-subject response latency standard deviation for each section (i.e., side [part I], direction [part II], switching sides [part IIIA] and switching directions [part IIIB] conditions) of the SWAT. 2.4. Procedure All study participants underwent nocturnal home or lab sleep monitoring immediately prior to their scheduled daytime performance testing. While undergoing these studies, participants were instructed to maintain customary bed times and wake times. Also, all home sleep studies were scheduled for nights when participants planned to have no overnight house guests. Those who used sleep medications on an intermittent basis were instructed to abstain from these medications for at least 2 weeks prior to and during nocturnal monitoring. Finally, all participants were instructed to abstain from alcoholic beverages and sedative medications entirely and to abstain from caffeinated substances after 1800 h on study nights. All performance trials were conducted in the sleep laboratory under the supervision of trained laboratory technologists. The first trial of the performance test battery began 2–3 h after the participant’s morning awakening, and successive trials were administered at 2-h intervals. Those who underwent LPSG prior to their performance testing remained in the sleep laboratory for this testing following their third LPSG night, whereas those who underwent HPSG prior to testing traveled to the laboratory for testing following their third HPSG study. Each performance trial was commenced at the instruction of the assigned technologist, and participants were supervised between trials to prevent unscheduled sleep episodes. After all participants completed the study, we combined data across testing trials by computing mean values for each of the reaction time latencies and within-subject standard deviation measures for each of the various reaction time subtests administered. We decided to combine data across testing trials because our primary objective was that of determining how our participants’ average daytime reaction time performances related to their amounts of slow-wave sleep during the night prior to testing. Once this was done, we computed correlations between the participants’ averaged response latencies and within-subject standard deviation scores for each performance subtest. Results of these analyses showed that response latencies and standard deviation scores for each test tended to be significantly correlated. Given these findings, we decided to retain the averaged response latency scores for our planned analyses so as to simplify our statistical procedures and data presentation. Two of our hypotheses required that we compare the results obtained from this study’s participants with both our [8] previous findings among older adults and the findings reported by Jurado et al. [14], who studied a young adult group. The research methodology used in both of these pre-
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
vious studies consisted of first dichotomizing study participants into Fast and Slow Performers based on their reaction time test performances, and then comparing these resultant groups in regard to measures of slow-wave sleep derived from the night before performance testing. To facilitate our comparisons with these previous studies, we used a similar methodological approach in the current investigation. As a first step, median response latency scores for each subtest were obtained for the female insomnia sufferers, the male insomnia sufferers, the female normal sleepers, and the male normal sleepers. These median scores were then used to dichotomize each of these four subgroups into Fast (i.e., those with averaged response latencies below the median) and Slow (i.e., those with averaged response latencies at or above the median) Performers. Furthermore, these dichotomies were formed for each of the six reaction time subtests. Subsequently the Stepwise Linear Regression procedure included in the SAS software package [19] was used to test if statistically normalized values of our slow-wave measures (i.e., SWS %, spectral analytically derived measures) predicted participants’ performance classification for each of the six reaction time tests within our samples of normals and PI sufferers. For the purpose of these prediction equations, a numerical value of “0” was used to designate Fast Performers, whereas a numerical value of “1” was used to designate Slow Performers. Because a set of preliminary analyses showed that the site where subjects slept during the night before performance testing had no effect on the regression results obtained, this factor was excluded from consideration in our final regression analyses. 3. Results 3.1. Descriptive data Table 1 shows the sleep and performance measures derived from our samples of normal sleepers and insomnia sufferers. Included in the table are means and standard deviations of various sleep parameters and reaction time measures as well as the results of statistical (ANOVA) comparisons of these two groups for each measure. When necessary, these data were normalized for the ANOVAs conducted. These analyses showed that the insomnia sufferers, on average, had significantly more wake time after sleep onset, lower sleep efficiencies, a lower percentage of slow-wave sleep and less power in the 0.5 to 2.0-Hz slow-wave bandwidth than did the normal sleepers (ps all ⬍ 0.05). In addition, the insomnia sufferers, as a group, performed significantly poorer on two SWAT subtests than did the normal sleepers. Comparisons using the remaining measures showed no significant differences between these two groups. 3.2. Slow-save sleep and performance within goups The regression analyses showed that spectral analytically derived slow-wave measures were predictive of normal sleepers’ performance classifications (i.e., Fast vs. Slow) for two of
131
Table 1 Means and standard deviations for standard and computer-derived sleep measures Measure
Normal sleepers
Measures of sleep time and consolidation Time in Bed (min) 441.1 (69.4) Total sleep time (min) 381.7 (54.0) Sleep onset (min) 14.2 (15.3) Wake after onset (min) 45.4 (43.4) Sleep efficiency % 87.8 (8.5)
Insomnia sufferers
F
p⬍
471.9 (42.9) 385.2 (42.9) 15.3 (9.6) 71.4 (35.4) 82.3 (8.1)
3.9 0.1 1.3 12.2 10.3
ns ns ns 0.001 0.003
78.0 (24.6) 16.6 (5.4) 63.2 (9.2) 15.8 (3.4)
2.8 6.3 6.9 1.9
ns 0.025 0.025 ns
229.5 (40.4) 362.7 (45.2) 280.4 (53.0) 464.5 (130.8) 543.2 (192.0) 689.8 (177.9)
0.5 0.7 3.4 4.8 3.0 4.1
ns ns ns .05 ns .05
Slow-Wave Measures Slow-wave sleep (min) Slow wave % % Power in the 0.5 to 2 Hz band % Power in the 2 to 4 Hz band
90.3 (30.2) 20.8 (6.9) 68.6 (6.3) 14.7 (2.8)
Performance measures—mean latencies (ms) Simple reaction time Continuous performance test SWAT—Part I SWAT—Part II SWAT—Part III-A—side SWAT—Part III-B—direction
222.0 (36.5) 353.8 (37.3) 257.7 (38.0) 406.7 (70.2) 459.7 (138.7) 611.9 (110.0)
Numbers in parentheses are standard deviations, whereas the other numbers are means; degrees of freedom for all F-tests are 1/56; ns ⫽ nonsignificant.
the six performances tests. The % power in the 0.5–2.0-Hz spectral band was found to be a modest (r 2 ⫽ 0.14), albeit significant, predictor, F(1, 29) ⫽ 4.70, p ⬍ 0.05, of performance on the Continuous Performance Test (CPT) (i.e., signal detection task), whereas the % power in the 2.0–4.0-Hz spectral band was a significant, F(1, 29) ⫽ 4.87, p ⬍ 0.05, albeit modest (r 2 ⫽ 0.14) predictor of performance on part II (i.e., directional responding condition) of the Switching Attention Test (SWAT-II). Figure 1 shows the nature of the relationship between these slow wave measures and performance on these two tests. Those categorized as Fast Performers on the CPT surprisingly had significantly less power in the 0.5–2.0-Hz slow-wave band than did the Slow Performers, but these two groups showed no significant differences in the power they displayed in the 2.0–4.0-Hz band. In contrast, Slow Performers on the SWAT-II had significantly more power in the 2.0– 4.0-Hz slow-wave band than did the Fast Performers, but these two groups did not differ statistically in terms of the % power they showed in the lower slow-wave band. Among the insomnia sufferers, neither spectral analytically derived slow-wave measure predicted performance on any of the six reaction time tasks. However, the percent of time these individuals spent in SWS on the night before testing was found to be a significant predictor of insomnia sufferers’ performances on the CPT, F(1, 26) ⫽ 6.25, p ⬍ 0.025, and Part IIIA (i.e., switching sides condition) of the SWAT, F(1, 26) ⫽ 8.01, p ⬍ 0.01. In the case of the CPT, slow wave % (SWS %) accounted for 20% (r2 ⫽ 0.20) of the between-group variance noted between Fast and Slow Performers, whereas in the case of the SWAT-IIIA (switching
132
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
Fig. 2. Relationship between slow-wave % and performance on the Continuous Performance Test and part III of the Switching Attention Test within the sample of isomnia sufferers..
Fig. 1. Relationship between spectral analytically derived slow-wave measures and performance on the Continuous Performance Test and part II of the Switching Attention Test within the sample of normal sleepers.
sides condition), this measure accounted for almost one-fourth (r 2 ⫽ 0.24) of the variance noted between these groups. Figure 2 shows the relationships found between SWS % and performance on these two tests within the insomnia group. This figure shows that Fast Performers on the CPT had a lower percentage of slow-wave sleep on the night before testing than did the Slow Performers on this test. In contrast, Fast Performers on part IIIA of the SWAT had a significantly higher percentage of nocturnal slow wave sleep than did the Slow Performers. Hence, as was the case in the normal sleeper sample, results derived from the PI sufferers showed that the relationship between slow-wave sleep and performance varied across the six reaction time tasks administered. 4. Discussion The current investigation was conducted to explore the relationship between nocturnal slow-wave sleep and subse-
quent diurnal reaction time among samples of middle-aged normal sleepers and insomnia sufferers. Consistent with the global hypothesis that slow wave sleep contributes to neurocognitive restoration, we predicted that middle-aged normal sleepers would show an inverse relationship between the amount/intensity of their nocturnal SWS and their subsequent diurnal reaction times. However, inasmuch as SWS tends to decline with aging, we predicted that this relationship would be less pronounced among our middle-aged cohort than that previously observed among young adult normal sleepers [14]. We also predicted that middle-aged PI sufferers would show a more pronounced SWS-reaction time performance relationship than would our age-matched normal sleepers. Unfortunately, our findings provided very limited support for our speculations, and suggested that the relationship between SWS and subsequent performance appears far more complex than we had predicted. Findings from both of our samples suggest that the relationship between nocturnal SWS and subsequent diurnal performance varies across performance tasks even when the measure of performance (i.e., response latency) remains constant. Indeed, findings from both the normal sleepers and insomnia sufferers showed that SWS measures were not predictive of performance on the Simple Reaction Time Test, nor on parts I and IIIB (switching directions condition) of the Switching Attention Test (SWAT). Data from both samples also showed that performance on the CPT signal detection task was related to slow-wave sleep, but not in the manner we had predicted. Among the normal sleepers, the better CPT performances had lower power in the 0.5–2.0Hz band than did the poorer performers, whereas among the insomnia sufferers, the better CPT performers had a lower SWS % than did the poorer performers. Also, seemingly paradoxical was the finding that those normal sleepers classified as Fast Performers on part II of the SWAT had lower power in the 2.0–4.0-Hz band than did those classified as Slow Performers. The only result consistent with our pre-
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
dictions was the finding that the insomnia sufferers classified as Fast Performers on part IIIA of the SWAT actually had a significantly higher SWS % on the night before testing than did those classified as Slow Performers. Overall, these findings seem to provide very limited support for the notion that SWS serves a restorative role in the neurocognitive functioning of middle-aged individuals. Although the reasons for the noted unexpected findings are not immediately obvious, it is possible that reaction time performance is dependent on such factors as the specific demands of the task in question and the endogenous arousal level of the respondent as well as the restorative value of the previous night’s sleep. For such relatively simple and subjectively boring tests as the CPT signal detection task, those individuals with relatively heightened endogenous states of arousal may be best able to maintain sufficient attention so as to achieve the best performances. If this assumption is correct, then indices such as relatively low power in the 0.5–2.0-Hz band (normal sleepers) or relatively low SWS % (insomnia sufferers) may identify those who are relatively hyperaroused and most able to maintain constant attention across unstimulating tasks. Conversely, more challenging tasks such as the subtests included in the SWAT may be more directly related to the restorative quality of the previous night’s SWS. Among normal sleep sleepers, those with relatively high power in the upper (2.0–4.0 Hz) band may derive a reduced restorative benefit from SWS, whereas a low SWS % may reflect a similar state among insomnia sufferers. If all of these assumptions are correct, then the findings we obtained seem understandable. However, additional studies that confirm our assumptions about the spectral analytically derived measures and include indices of endogenous arousal will be needed to confirm these speculations. Despite the noted unexpected findings, our results did support our speculation that the nocturnal SWS–diurnal performance relationship among PI sufferers is different from that seen among normal sleepers. Perhaps due to the fact that they evidenced lower power overall in the spectral analytically derived measures than did the normal sleepers, the insomnia sufferers showed no relationship between these supposedly more sensitive slow-wave measures and their performances on any of the tests administered. However, they did show significant relationships between their SWS % derived from conventional scoring methods [18] and performances on two of the performance tests. Moreover, r 2 values derived from prediction equations suggested that these relationships were somewhat stronger than the SWS–performance relationships observed among the normal sleepers. This observation and the findings noted in our previous study (8) are consistent with our speculation that the presence of sleep pathology such as PI may actually potentiate the observed association between nocturnal SWS and subsequent diurnal performance. Thus, future studies to explore mechanisms (e.g., sleep variability; diurnal hyperarousal) potentially responsible for this alteration would be useful.
133
In addition to these various observations, it seems useful to compare our past [8] and current findings with those reported by Jurado et al. [14]. Whereas Jurado et al. [14] found a strong inverse relationship between nocturnal slowwave sleep and subsequent diurnal reaction time among young normal sleepers, we failed to replicate this finding among middle-aged and older normal sleepers in our current and previous studies. Data derived from our middleaged cohort suggested significant but different SWS–performance relationships among normal sleepers than those found by Jurado et al. [14], whereas data from our [8] senior cohort failed to show any SWS–performance relationships. We should note that these various studies differed methodologically in that Jurado et al. compared SWS of the fastest and slowest thirds of their sample, whereas we used a median split to divide our participants into Fast and Slow Performers in our two studies so as to maximize our statistical power. Nonetheless, the cross-sectional comparisons these studies allow provides some support for the notion [21] that the restorative role of SWS at least for simple neurocognitive functions like reaction time, may diminish across the lifespan. Of course, longitudinal studies would be useful to provide further confirmation for this hypothesis. Admittedly, the current investigation had a number of limitations that warrant consideration. Our sample included normal sleepers and PI sufferers who, by and large, presented to us as research volunteers. Whether our findings apply to middle-aged normals in general and clinical samples of PI patients remains to be determined. It also should be noted that our computer software was calibrated to score a very limited EEG frequency range. As a result, our analyses of these measures connoted differences only in relative power rather than the absolute or total power manifest across the entire EEG spectrum. Given our speculations about endogenous arousal, it may have been useful to obtain power data for those higher frequency bands (e.g., beta, alpha, etc.), which might reflect such arousal. In addition, it should be mentioned that we assessed our subjects’ daytime performance via a limited number of reaction time tasks. As a result, it remains possible that other dimensions of cognitive functioning such as memory, language, reasoning, etc., may show relationships to slow-wave sleep in both the age group studied and in other age groups. Finally, our research approach involving between-subjects comparisons of Fast and Slow performers may have prevented us from discerning the actual importance of SWS to diurnal performance. Perhaps within-subjects comparisons of study participants following both normal/good nights of sleep and nights during with their SWS is experimentally or naturally interrupted would provide somewhat different results than those we obtained. Thus, future studies would benefit by the inclusion of clinical patients, larger research samples, more diverse age groups, a more comprehensive scoring of the EEG spectrum, a wider range of performance measures, and the use of within-subjects methodologies so as to provide a better understanding of importance of SWS to neurocognitive functioning.
134
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
Acknowledgment This research was supported by the Department of Veterans Affairs Merit Review Program. References [1] American Psychiatric Press. Structured Clinical Interview for DSMIII-R (SCID). Washington, DC: American Psychiatric Press, 1990. [2] Armitage R, Hoffman R, Fitch T. A comparison of period-amplitude and spectral analysis in quantifying sleep EEG: normal controls. Boston, MA: Annual Meeting of the Associated Professional Sleep Societies, 1994. [3] Berry DTR, Webb WB. Sleep and cognitive functions in normal older adults. J Gerontol 1985;40:331. [4] Bliwise DL. Sleep in normal aging and dementia. Sleep 1993;16:40–81. [5] Bonnet MH, Arand DL. 24-hour metabolic rate in insomniacs and matched normal sleepers. Sleep 1995;18:581–8. [6] Bootzin RR. Effects of self-control procedures for insomnia. In Stuart RB, editor. Behavioral Self-Management. New York: Brunner/Mazel, Inc., 1977. [7] Bunnell DE, Bevier W, Horwath SM. Effects of exhaustive exercise on the sleep of men and women. Psychophysiology 1983;20:50–58. [8] Crenshaw MC, Edinger JD. Slow wave sleep and waking cognitive performance among older adults with and without insomnia complaints. Physiol Behav 1999;66:485–92. [9] Edinger JD, Wohlgemuth WK. The significance and management of persistent primary insomnia. Sleep Med Rev 1999;3:101–18. [10] Edinger JD, Morey MC, Sullivan RJ, Higginbotham MB, Marsh GR, Dailey DS, McCall WV. Aerobic fitness, acute exercise and sleep in older men. Sleep 1993;16:351–9. [11] Feinberg I, Koresko RL, Heller N, et al. EEG sleep patterns as a function of normal and pathological aging in man. J Psychiatr Res 1967;5:107. [12] Freedman RR. EEG power spectra in sleep-onset insomnia. Electroencephalogr Clin Neurophysiol 1986;63:408–13.
[13] Honda Y, Takahashi K, Takahashi S, Azumi K, Irie M, Sakuma M, Tsushima T, Schizume K. Growth hormone secretion during nocturnal sleep in normal subjects. J Clin Endocrinol Metab 1969; 29. [14] Jurado JL, Luna Villegas G, Buela–Casal G. Normal human subjects with slow reaction times and larger time estimations after waking have diminished delta sleep. Electroencephalogr Clin Neurophysiol 1989;73:124–8. [15] Letz R, Baker EL. NES2 Neurobehavioral Evaluation System. Chapel Hill: University of North Carolina, 1988. [16] Perlis ML, Giles DE, Mendelson WB, Bootzin RR, Wyatt JK. Psychophysiological insomnia: the behavioural model and a neurocognitive perspective. J Sleep Res 1997;6:179–88. [17] Prinz PN. Sleep patterns in healthy aged: relationship with intellectual function. J Gerentol 1997;32:179. [18] Rechtshaffen A, Kales A. A Manual of Standardized Terminology, Techniques, and Scoring Systems of Sleep Stages of Human Subjects. Los Angeles: UCLA Brain Information Service/Brain Research Institute, 1968. [19] SAS Institute, Inc. SAS Language and Procedures: Usage, ver. 6, 1st ed. Cary, NC: SAS Institute, Inc., 1989. [20] Schramm E, Hohagen F, Grasshoff U, Berger M. Test-retest reliability and validity of the Structured Interview for Sleep Disorders according to DSM-III-R. Am J Psychiatry 1993;150:867–72. [21] Spiegel R, Koberle S, Allen SR. Significance of slow wave sleep: considerations from a clinical viewpoint. Sleep 1986;9:66–79. [22] Spielman AJ, Saskin P, Thorpy MJ. Treatment of chronic insomnia by restriction of time in bed. Sleep 1987;10:45–55. [23] Takahashi Y, Kipnis D, Daughaday W. Growth hormone secretion during sleep. J Clin Investigat 1968;47:2079–90. [24] Tilley AJ, Wilkinson RT. The effects of a restricted sleep regime on the composition of sleep and on performance. Psychophysiology 1984;21:406–12. [25] Vitiello MV, Schwartz RS, Davis MW, Ward RR, Ralph DD, Prinz PN. Sleep quality and increased aerobic fitness in healthy aged men: preliminary findings. J Sleep Res 1992;1(Suppl. 1):245.
Slow-wave sleep and waking cognitive performance II: Findings among middle-aged adults with and without insomnia complaints Jack D. Edingera,b,*, D. Michael Glennc,d, Lori A. Bastiana,b, Gail R. Marshb a
Veterans Affairs and bDuke University Medical Centers, Durham, NC, USA c John Umstead Hospital, Butner, NC, USA d University of North Carolina, Chapel Hill, NC, USA Received 9 June 1999; received in revised form 8 November 1999; accepted 10 February 2000
Abstract Previous studies showing a relationship between nocturnal slow-wave sleep (SWS) and subsequent diurnal performance among young normal sleepers and older insomnia sufferers have provided limited support for the notion that this sleep stage serves a restorative role for neurocognitive functioning. The current study, which examined the relationship between SWS and reaction time performance among middle-aged adults with and without insomnia complaints, was conducted to further explore this possibility. A sample of 31 noncomplaining middle-aged (ages 40 to 59 years) normal sleepers and a like-aged sample of 27 insomnia sufferers, provided data for the current investigation. All participants underwent nocturnal sleep monitoring immediately prior to undergoing a battery of daytime tests that measured simple reaction time, vigilance/signal detection, and complex reaction time. Results showed relationships between reaction time performances on some tasks and some SWS measures among both the normal sleepers and insomnia sufferers. Findings supported our prediction that the presence of sleep pathology (e.g., insomnia) alters the SWS–performance relationship observed, but the results failed to show a consistent relationship between SWS and subsequent performance within either sample. The findings suggest that the specific performance demands of the task in question as well as physiological parameters other than SWS may determine performance as well. Findings for this and previous studies do provide some support for the contention that the neurocognitive restorative value of SWS may change across the lifespan. Possible implications of the study’s findings are discussed and directions for future research are considered. © 2000 Elsevier Science Inc. All rights reserved. Keywords: Slow-wave sleep; Waking cognitive performance; Middle-aged adults
1. Introduction Numerous studies [7,10,13,23–25] of sleep deprivation, growth hormone secretion, acute exercise, and physical fitness have implied that slow-wave or deep sleep may contribute to somatic restoration among various age groups. Although it has been theorized that nocturnal slow-wave sleep (SWS) might also serve a restorative role for neurocognitive functions, previous studies exploring the relationship between SWS and diurnal performance [3,8,11,14,17,21] have provided mixed results. Perhaps this variability is understandable inasmuch as the complexity of neurocognitive tasks employed in these studies has varied appreciably. Nonetheless, even those studies using similar performance tasks have produced inconsistent findings. Jurado et al. [14], * Corresponding author. VA Medical Center, Psychology Service (116B), 508 Fulton Street, Durham, NC 27705. Tel.: (919) 286-0411 ext. 17054; Fax: (919) 416-5832. E-mail address: [email protected].
for example, found a significant relationship between the amounts of SWS that young healthy adults obtained at night and their subsequent diurnal reaction times. In contrast, we [8] failed to replicate this finding among older (age ⭓60 years) adults with and without insomnia complaints. However, using slow-wave measures derived from computerassisted spectral analyses, we found that only insomnia sufferers showed an inverse relationship between daytime reaction time and EEG power in the 2 to 4-Hz spectral band. Hence, the potential restorative value of SWS for even fairly simple neurocognitive functions like reaction time remains questionable. Of course, one hypothesis suggested by these latter studies is that the degree to which slow-wave sleep restores neurocognitive processes among normal sleepers changes as a function of aging. Although Jurado et al. [14] found a strong relationship between nocturnal SWS and subsequent diurnal performance in healthy young normal sleepers, we and others failed to show a similar relation-
0031-9384/00/$ – see front matter © 2000 Elsevier Science Inc. All rights reserved. PII: S0031-9384(00)00 2 3 8 - 9
128
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
ship among senior adults. This observation as well as the fact that normal sleepers show marked declines in SWS across the lifespan [1,4] suggests that the degree to which SWS contributes to neurocognitive restoration may gradually decline through adulthood. Indeed, by late life, SWS may become what some investigators have called a “functionally meaningless remnant” [21]. Unfortunately there currently is no longitudinal data to corroborate this speculation. Furthermore, because previous studies of the SWS– neurocognitive performance relationship have generally ignored middle-aged normal sleepers, there currently is insufficient published cross-sectional data to support this contention. Thus, studies of the SWS–performance relationship among middle-aged normal sleepers would help fill this void in this literature. In addition, our previous findings suggest that conditions such as chronic primary insomnia (PI), may alter the observed relationship between SWS and neurocogntive functioning. As we and others [6,9,22] have noted, PI is typically perpetuated by aberrant sleep habits and conditioned arousal, which interfere with normal homeostatic regulation of the sleep process. As a consequence, those who suffer from PI experience enhanced night-to-night variability in SWS and other sleep parameters. Indeed, such individuals typically show sleep patterns characterized by one or more nights of poor sleep (i.e., partial deprivation), followed by one or more nights of good or extended sleep (i.e., recovery). Despite or, perhaps, due to their chronically unpredictable sleep, PI sufferers often appear hyperalert or neuocognitively aroused during the daytime [5,16]. The combination of erratic nocturnal sleep and diurnal hyperalertness/arousal could arguably alter the SWS–diurnal performance relationship that is observed among like-aged normal sleepers. Whereas our previous study with older adults suggests this may be the case, additional tests of this speculation seem warranted. The current investigation was conducted, in part, to explore the possible restorative role of nocturnal SWS for subsequent neurocognitive, reaction time functioning among middle-aged normal sleepers. In addition, this study was conducted to replicate our previous findings, which suggested that the nocturnal SWS–diurnal performance relationship observed in these normal sleepers varies from that observed within an age-matched sample of insomnia sufferers. To accomplish these study objectives, we extracted sleep and reaction time performance data obtained in a larger study of middle-aged adults with and without insomnia complaints. For the purposes of this study, we hypothesized that: (1) middle-aged normal sleepers would show an inverse relationship between the amount and intensity of their nocturnal SWS and their subsequent diurnal reaction times; (2) this SWS–reaction time relationship would be less pronounced among these middle-aged normal sleepers than that previously observed among young adult normal sleepers (e.g. [14]); and (3) middle-aged PI sufferers would show a more pronounced SWS–reaction
time performance relationship than would our age-matched normal sleepers.
2. Materials and methods 2.1. Subjects Data gathered for the current investigation were obtained from a cohort of middle-aged (40 to 59 years) adults while they participated in a larger study designed to address a number of research objectives in addition to those pertinent to the current report. Participants included in this study were drawn from samples of 32 (16 women, 16 men) middleaged normal sleepers and 32 age- and gender-matched insomnia sufferers who were recruited through announcements posted within two medical centers (i.e., Durham VA and Duke University Medical Centers) and via recruitment letters sent to Durham VA Medical Center volunteer workers and to individuals listed in the subject pool of the Duke Center for the Study of Aging and Human Development. The normal sleepers had no acute, sleep-disruptive medical disorders, reported no sleep complaints, and did not meet criteria for any sleep disorder on the basis of a Structured Interview for Sleep Disorders (SIS-D) [20]. The insomnia sufferers were also medically healthy individuals who: (a) had a history of insomnia complaints 艌6 months duration; (b) met SIS-D criteria for Primary Insomnia (PI); and (c) did not meet SIS-D criteria for any other sleep disorder. Excluded from the study were individuals who: (a) had a terminal illness; (b) had an unstable medical condition or a medical disorder (e.g., rheumatoid arthritis, thyroid disease, asthma) that compromises sleep; (c) had abnormal TSH levels on a screening thyroid panel (d) had a history of psychiatric illness(es); (e) met criteria for a major psychiatric (Axis I) condition on the basis of a Structured Clinical Interview for psychiatric disorders (SCID) (1); (f) were habitual substance abusers; (g) used sleep medications and were unwilling/unable to abstain from this practice; (h) were taking anxiolytics, antidepressants, or any other psychotropic medication; or (i) had 艌15 events per hour characterized by a total cessation of breathing (i.e., apneas) and/or dramatic reduction in airflow (i.e., hypopneas ⫽ 50% or greater decrease in airflow) on a screening night of sleep monitoring. Of all individuals enrolled in our research program, three participants (two male normal sleepers and one female insomnia sufferer) were dropped from consideration for the current investigation because their sleep recordings for the night before performance testing were of such poor technical quality that they were deemed unscoreable. We also noted that some performance data were lost from three male insomnia sufferers as a result of technician error in storing these data into computer memory. As a result, each of these individuals had missing performance data for some of the testing trials. Because our experience suggests that individuals tend to show both practice and circadian fluctuations in their responses to the performances tests used in this study,
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
we assumed that averaged performance data taken from individuals with missing data for some trials might not accurately represent their neurocognitive functioning across the day. Given this rationale, data from these three participants with missing data were eliminated from consideration in the analyses conducted in this investigation. Our remaining sample consisted of 31 (16 women, 15 men) normal sleepers (mean age ⫽ 46.4 years, SD ⫽ 5.2 years) and 27 (15 women, 12 men) PI sufferers (mean age ⫽ 49.7 years, SD ⫽ 5.7 years). As a group, those with insomnia had, on average, completed 15.3 (SD ⫽ 2.8 years) years of education and had had subjective sleep complaints for an average of 10.5 years (SD ⫽ 7.8 years). Three (5.2%) of these participants complained exclusively of sleep-onset difficulties, 12 (44.4%) had solely sleep-maintenance complaints, 11 (40.7%) reported a mixture of sleep onset and maintenance problems, and the remaining subject reported problems with chronically poor sleep quality. At the time of their study participation, 20 (74.1%) of the insomnia sufferers reported no current use of sedative hypnotics (e.g., benzodiazepines, zolpidem, etc.), 3 (11.1%) reported use of such agents one time per week or less often, 1 (3.7%) used such agents less than two times per week, and the remaining 3 (11.1%) used such agents two to four times per week to aid their sleep. The normal sleepers had comparable educational histories (mean years of education ⫽ 16.3 years; SD ⫽ 2.6 years) but reported no sleep difficulties and no current use of hypnotic sleep medications. For the purposes of this study, all participants using sedative hypnotics abstained from these medications during a 2-week wash-out period as well as during their participation in the study. 2.2. Polysomnographic sleep monitoring All sleep monitoring studies were conducted with Oxford Medilog 9000/9200 (Oxford Medical, Inc., Clearwater, FL) ambulatory recording devices. The Medilog is a cassette system with the capability of recording eight channels of electrophysiological data in analog format, as well as digital time in 1-s intervals, and subject-labeled events. Our monitoring montage for the current study included two EEG channels (C3-M2, Oz-Cz), one chin EMG channel, two channels to monitor eye movements, EOG (left eye—M1, right eye—M2), and two channels to monitor anterior tibialis EMG (right and left legs). In addition, respiration (nasal/ oral air flow–thermistor) was monitored to screen for apneic events. For purposes of the above-mentioned larger study, all individuals included in the current study underwent a total of 6 nights of sleep recording with either a series of three home (HPSG) or three lab (LPSG) recordings occurring initially in counterbalanced order. For all study participants, daytime performance testing occurred after the first series of three PSG studies (either HPSG or LPSG). By random assignment, one-half (16 normals, 16 insomniacs) of the 64 participants enrolled underwent their three HPSG studies prior to daytime testing, whereas the other half of the sample underwent three LPSGs prior to daytime testing. Of
129
these, the 58 who met selection criteria for the current study consisted of 30 (15 normal sleepers, 15 insomnia sufferers) who underwent HPSG prior to daytime testing and 28 (16 normal sleepers, 12 insomnia sufferers) who underwent LPSG prior to this testing. Once these individuals had completed all experimental procedures, sleep data from the night of monitoring immediately preceding daytime testing (i.e., third night) were extracted for the purposes of this study. Because previous studies [2,12] and our own experience [8] have suggested that computer-assisted EEG spectral analyses may provide a more sensitive characterization of slow-wave sleep than do conventional manual scoring techniques, we used this method in addition to conventional manual scoring techniques [18] to score our participants’ sleep recordings. For each recording, a Fourier transformation of the sleep data and subsequent spectral analysis was used to quantify the: (a) % of EEG power within the 0.5 to 2-Hz bandwidth during NREM sleep; and (b) % of EEG power within the 2 to 4-Hz bandwidth during NREM sleep. The computer software used was designed to exclude periods of wakefulness, stage 1, and REM sleep as well as movement artifacts from the two spectral measures derived. These spectral analytically derived measures along with estimates of SWS % (i.e., [SWS-min/time in bed-min] ⫻ 100%) derived from conventional scoring methods [18] served as the primary SWS measures used to test the study hypotheses. 2.3. Performance testing All study participants completed four 16-min trials composed of a series of three computer-administered reaction time tasks selected from the Neurobehavioral Evaluation System (NES) [15]. During the test administration, each participant was placed individually in a testing room, and was seated in front of a PC computer that contained the NES software used for the testing. At the beginning of each test, the computer software provided written instructions on the computer screen. The study participant read the instructions and then proceeded with the computer-guided standardized battery as outlined in the ensuing discussion. 2.3.1. Simple Reaction Time Test The first and least difficult test presented in each trial was the Simple Reaction Time Test (SRT). During this test, the participant was required to press a specially marked key on a computer keyboard whenever a figure (i.e., a small square) appeared on the computer screen. Once the test was begun, the figure appeared at intervals varying between 1000 and 2500 ms. Upon each presentation, the figure remained on the screen either until a response occurred or 1000 ms had elapsed. The total trial lasted approximately 5 min, and consisted of 90 presentations of the target stimulus. For each presentation of the figure, the computer software automatically recorded the time (in ms) between the appearance of the figure on the screen and the computer key-press response. For each 5-min SRT trial, the computer software
130
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
computed a mean response latency and a within-subject standard deviation of the participant’s response latencies. 2.3.2. Continuous Performance Test This somewhat more challenging test consisted of a signal detection task during which a target (i.e., the letter “S”) and background signals (i.e., the letters “A,” “C,” “E,” and “T”) were presented in rapid fashion on the computer screen. Once the test was begun, target and background letters appeared in a random sequence during which a 1:4 target-to-background ratio was maintained. Target and background letters were presented at the rate of 1 per second, and each letter remained on the computer screen for 50 ms. The total test lasted approximately 5 min, and included 60 presentations of the target and 240 presentations of the background stimuli. In responding to this test, the participant was required to press a specially marked key on the computer keyboard when and only when the preidentified target letter appeared on the screen. For each trial of this test, the NES software derived a mean response latency and a within-subject standard deviation for the response latencies. 2.3.3. Switching Attention Test The final test, the Switching Attention Test (SWAT), lasted approximately 6 min, and included three subtests. Each subtest required the participant to press specially marked keys on the right and left side of the keyboard in response to stimuli presented on the computer screen. During part I (side condition) of the SWAT, a square appeared either on the right or left side of the computer screen and the participant was required to press a marked key on the corresponding side of the computer keyboard. Upon each presentation, the stimulus remained on the screen either until a response occurred or 2500 ms had elapsed. This portion of the SWAT included 6 practice and 16 test presentations of the stimulus. During part II (direction condition), an arrow, pointing right or left, appeared in the center of the screen, and the participant was required to make right or left side key presses in response to the direction in which the arrow was pointing. As in the previous section of the SWAT, the stimulus remained on the screen either until a response occurred or 2500 ms had elapsed. This portion of the test included 4 practice and 16 test presentations of the stimulus. Finally, during parts IIIA and IIIB (switching conditions), an arrow (pointing right or left) appeared either on the right or left side of the screen. Preceding each presentation of this arrow by 1000 ms, one of two command words, “SIDE” or “DIRECTION,” appeared on the screen. This command word served to instruct the participant to respond by pressing a key on the side of the keyboard corresponding either to the side of the screen on which the arrow appeared or the direction in which the arrow was pointing. On 50% of the presentations the side of the screen on which the arrow appeared and the direction in which it was pointing agreed. On the remaining presentations, these two stimulus characteristics were in conflict. Throughout the test, these nonconflict and conflict presentations occurred in a random se-
quence. Overall, the switching condition included 8 practice and 48 test presentations of the command-stimulus combination. For each test trial, the NES computer software computed a mean response latency and a within-subject response latency standard deviation for each section (i.e., side [part I], direction [part II], switching sides [part IIIA] and switching directions [part IIIB] conditions) of the SWAT. 2.4. Procedure All study participants underwent nocturnal home or lab sleep monitoring immediately prior to their scheduled daytime performance testing. While undergoing these studies, participants were instructed to maintain customary bed times and wake times. Also, all home sleep studies were scheduled for nights when participants planned to have no overnight house guests. Those who used sleep medications on an intermittent basis were instructed to abstain from these medications for at least 2 weeks prior to and during nocturnal monitoring. Finally, all participants were instructed to abstain from alcoholic beverages and sedative medications entirely and to abstain from caffeinated substances after 1800 h on study nights. All performance trials were conducted in the sleep laboratory under the supervision of trained laboratory technologists. The first trial of the performance test battery began 2–3 h after the participant’s morning awakening, and successive trials were administered at 2-h intervals. Those who underwent LPSG prior to their performance testing remained in the sleep laboratory for this testing following their third LPSG night, whereas those who underwent HPSG prior to testing traveled to the laboratory for testing following their third HPSG study. Each performance trial was commenced at the instruction of the assigned technologist, and participants were supervised between trials to prevent unscheduled sleep episodes. After all participants completed the study, we combined data across testing trials by computing mean values for each of the reaction time latencies and within-subject standard deviation measures for each of the various reaction time subtests administered. We decided to combine data across testing trials because our primary objective was that of determining how our participants’ average daytime reaction time performances related to their amounts of slow-wave sleep during the night prior to testing. Once this was done, we computed correlations between the participants’ averaged response latencies and within-subject standard deviation scores for each performance subtest. Results of these analyses showed that response latencies and standard deviation scores for each test tended to be significantly correlated. Given these findings, we decided to retain the averaged response latency scores for our planned analyses so as to simplify our statistical procedures and data presentation. Two of our hypotheses required that we compare the results obtained from this study’s participants with both our [8] previous findings among older adults and the findings reported by Jurado et al. [14], who studied a young adult group. The research methodology used in both of these pre-
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
vious studies consisted of first dichotomizing study participants into Fast and Slow Performers based on their reaction time test performances, and then comparing these resultant groups in regard to measures of slow-wave sleep derived from the night before performance testing. To facilitate our comparisons with these previous studies, we used a similar methodological approach in the current investigation. As a first step, median response latency scores for each subtest were obtained for the female insomnia sufferers, the male insomnia sufferers, the female normal sleepers, and the male normal sleepers. These median scores were then used to dichotomize each of these four subgroups into Fast (i.e., those with averaged response latencies below the median) and Slow (i.e., those with averaged response latencies at or above the median) Performers. Furthermore, these dichotomies were formed for each of the six reaction time subtests. Subsequently the Stepwise Linear Regression procedure included in the SAS software package [19] was used to test if statistically normalized values of our slow-wave measures (i.e., SWS %, spectral analytically derived measures) predicted participants’ performance classification for each of the six reaction time tests within our samples of normals and PI sufferers. For the purpose of these prediction equations, a numerical value of “0” was used to designate Fast Performers, whereas a numerical value of “1” was used to designate Slow Performers. Because a set of preliminary analyses showed that the site where subjects slept during the night before performance testing had no effect on the regression results obtained, this factor was excluded from consideration in our final regression analyses. 3. Results 3.1. Descriptive data Table 1 shows the sleep and performance measures derived from our samples of normal sleepers and insomnia sufferers. Included in the table are means and standard deviations of various sleep parameters and reaction time measures as well as the results of statistical (ANOVA) comparisons of these two groups for each measure. When necessary, these data were normalized for the ANOVAs conducted. These analyses showed that the insomnia sufferers, on average, had significantly more wake time after sleep onset, lower sleep efficiencies, a lower percentage of slow-wave sleep and less power in the 0.5 to 2.0-Hz slow-wave bandwidth than did the normal sleepers (ps all ⬍ 0.05). In addition, the insomnia sufferers, as a group, performed significantly poorer on two SWAT subtests than did the normal sleepers. Comparisons using the remaining measures showed no significant differences between these two groups. 3.2. Slow-save sleep and performance within goups The regression analyses showed that spectral analytically derived slow-wave measures were predictive of normal sleepers’ performance classifications (i.e., Fast vs. Slow) for two of
131
Table 1 Means and standard deviations for standard and computer-derived sleep measures Measure
Normal sleepers
Measures of sleep time and consolidation Time in Bed (min) 441.1 (69.4) Total sleep time (min) 381.7 (54.0) Sleep onset (min) 14.2 (15.3) Wake after onset (min) 45.4 (43.4) Sleep efficiency % 87.8 (8.5)
Insomnia sufferers
F
p⬍
471.9 (42.9) 385.2 (42.9) 15.3 (9.6) 71.4 (35.4) 82.3 (8.1)
3.9 0.1 1.3 12.2 10.3
ns ns ns 0.001 0.003
78.0 (24.6) 16.6 (5.4) 63.2 (9.2) 15.8 (3.4)
2.8 6.3 6.9 1.9
ns 0.025 0.025 ns
229.5 (40.4) 362.7 (45.2) 280.4 (53.0) 464.5 (130.8) 543.2 (192.0) 689.8 (177.9)
0.5 0.7 3.4 4.8 3.0 4.1
ns ns ns .05 ns .05
Slow-Wave Measures Slow-wave sleep (min) Slow wave % % Power in the 0.5 to 2 Hz band % Power in the 2 to 4 Hz band
90.3 (30.2) 20.8 (6.9) 68.6 (6.3) 14.7 (2.8)
Performance measures—mean latencies (ms) Simple reaction time Continuous performance test SWAT—Part I SWAT—Part II SWAT—Part III-A—side SWAT—Part III-B—direction
222.0 (36.5) 353.8 (37.3) 257.7 (38.0) 406.7 (70.2) 459.7 (138.7) 611.9 (110.0)
Numbers in parentheses are standard deviations, whereas the other numbers are means; degrees of freedom for all F-tests are 1/56; ns ⫽ nonsignificant.
the six performances tests. The % power in the 0.5–2.0-Hz spectral band was found to be a modest (r 2 ⫽ 0.14), albeit significant, predictor, F(1, 29) ⫽ 4.70, p ⬍ 0.05, of performance on the Continuous Performance Test (CPT) (i.e., signal detection task), whereas the % power in the 2.0–4.0-Hz spectral band was a significant, F(1, 29) ⫽ 4.87, p ⬍ 0.05, albeit modest (r 2 ⫽ 0.14) predictor of performance on part II (i.e., directional responding condition) of the Switching Attention Test (SWAT-II). Figure 1 shows the nature of the relationship between these slow wave measures and performance on these two tests. Those categorized as Fast Performers on the CPT surprisingly had significantly less power in the 0.5–2.0-Hz slow-wave band than did the Slow Performers, but these two groups showed no significant differences in the power they displayed in the 2.0–4.0-Hz band. In contrast, Slow Performers on the SWAT-II had significantly more power in the 2.0– 4.0-Hz slow-wave band than did the Fast Performers, but these two groups did not differ statistically in terms of the % power they showed in the lower slow-wave band. Among the insomnia sufferers, neither spectral analytically derived slow-wave measure predicted performance on any of the six reaction time tasks. However, the percent of time these individuals spent in SWS on the night before testing was found to be a significant predictor of insomnia sufferers’ performances on the CPT, F(1, 26) ⫽ 6.25, p ⬍ 0.025, and Part IIIA (i.e., switching sides condition) of the SWAT, F(1, 26) ⫽ 8.01, p ⬍ 0.01. In the case of the CPT, slow wave % (SWS %) accounted for 20% (r2 ⫽ 0.20) of the between-group variance noted between Fast and Slow Performers, whereas in the case of the SWAT-IIIA (switching
132
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
Fig. 2. Relationship between slow-wave % and performance on the Continuous Performance Test and part III of the Switching Attention Test within the sample of isomnia sufferers..
Fig. 1. Relationship between spectral analytically derived slow-wave measures and performance on the Continuous Performance Test and part II of the Switching Attention Test within the sample of normal sleepers.
sides condition), this measure accounted for almost one-fourth (r 2 ⫽ 0.24) of the variance noted between these groups. Figure 2 shows the relationships found between SWS % and performance on these two tests within the insomnia group. This figure shows that Fast Performers on the CPT had a lower percentage of slow-wave sleep on the night before testing than did the Slow Performers on this test. In contrast, Fast Performers on part IIIA of the SWAT had a significantly higher percentage of nocturnal slow wave sleep than did the Slow Performers. Hence, as was the case in the normal sleeper sample, results derived from the PI sufferers showed that the relationship between slow-wave sleep and performance varied across the six reaction time tasks administered. 4. Discussion The current investigation was conducted to explore the relationship between nocturnal slow-wave sleep and subse-
quent diurnal reaction time among samples of middle-aged normal sleepers and insomnia sufferers. Consistent with the global hypothesis that slow wave sleep contributes to neurocognitive restoration, we predicted that middle-aged normal sleepers would show an inverse relationship between the amount/intensity of their nocturnal SWS and their subsequent diurnal reaction times. However, inasmuch as SWS tends to decline with aging, we predicted that this relationship would be less pronounced among our middle-aged cohort than that previously observed among young adult normal sleepers [14]. We also predicted that middle-aged PI sufferers would show a more pronounced SWS-reaction time performance relationship than would our age-matched normal sleepers. Unfortunately, our findings provided very limited support for our speculations, and suggested that the relationship between SWS and subsequent performance appears far more complex than we had predicted. Findings from both of our samples suggest that the relationship between nocturnal SWS and subsequent diurnal performance varies across performance tasks even when the measure of performance (i.e., response latency) remains constant. Indeed, findings from both the normal sleepers and insomnia sufferers showed that SWS measures were not predictive of performance on the Simple Reaction Time Test, nor on parts I and IIIB (switching directions condition) of the Switching Attention Test (SWAT). Data from both samples also showed that performance on the CPT signal detection task was related to slow-wave sleep, but not in the manner we had predicted. Among the normal sleepers, the better CPT performances had lower power in the 0.5–2.0Hz band than did the poorer performers, whereas among the insomnia sufferers, the better CPT performers had a lower SWS % than did the poorer performers. Also, seemingly paradoxical was the finding that those normal sleepers classified as Fast Performers on part II of the SWAT had lower power in the 2.0–4.0-Hz band than did those classified as Slow Performers. The only result consistent with our pre-
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
dictions was the finding that the insomnia sufferers classified as Fast Performers on part IIIA of the SWAT actually had a significantly higher SWS % on the night before testing than did those classified as Slow Performers. Overall, these findings seem to provide very limited support for the notion that SWS serves a restorative role in the neurocognitive functioning of middle-aged individuals. Although the reasons for the noted unexpected findings are not immediately obvious, it is possible that reaction time performance is dependent on such factors as the specific demands of the task in question and the endogenous arousal level of the respondent as well as the restorative value of the previous night’s sleep. For such relatively simple and subjectively boring tests as the CPT signal detection task, those individuals with relatively heightened endogenous states of arousal may be best able to maintain sufficient attention so as to achieve the best performances. If this assumption is correct, then indices such as relatively low power in the 0.5–2.0-Hz band (normal sleepers) or relatively low SWS % (insomnia sufferers) may identify those who are relatively hyperaroused and most able to maintain constant attention across unstimulating tasks. Conversely, more challenging tasks such as the subtests included in the SWAT may be more directly related to the restorative quality of the previous night’s SWS. Among normal sleep sleepers, those with relatively high power in the upper (2.0–4.0 Hz) band may derive a reduced restorative benefit from SWS, whereas a low SWS % may reflect a similar state among insomnia sufferers. If all of these assumptions are correct, then the findings we obtained seem understandable. However, additional studies that confirm our assumptions about the spectral analytically derived measures and include indices of endogenous arousal will be needed to confirm these speculations. Despite the noted unexpected findings, our results did support our speculation that the nocturnal SWS–diurnal performance relationship among PI sufferers is different from that seen among normal sleepers. Perhaps due to the fact that they evidenced lower power overall in the spectral analytically derived measures than did the normal sleepers, the insomnia sufferers showed no relationship between these supposedly more sensitive slow-wave measures and their performances on any of the tests administered. However, they did show significant relationships between their SWS % derived from conventional scoring methods [18] and performances on two of the performance tests. Moreover, r 2 values derived from prediction equations suggested that these relationships were somewhat stronger than the SWS–performance relationships observed among the normal sleepers. This observation and the findings noted in our previous study (8) are consistent with our speculation that the presence of sleep pathology such as PI may actually potentiate the observed association between nocturnal SWS and subsequent diurnal performance. Thus, future studies to explore mechanisms (e.g., sleep variability; diurnal hyperarousal) potentially responsible for this alteration would be useful.
133
In addition to these various observations, it seems useful to compare our past [8] and current findings with those reported by Jurado et al. [14]. Whereas Jurado et al. [14] found a strong inverse relationship between nocturnal slowwave sleep and subsequent diurnal reaction time among young normal sleepers, we failed to replicate this finding among middle-aged and older normal sleepers in our current and previous studies. Data derived from our middleaged cohort suggested significant but different SWS–performance relationships among normal sleepers than those found by Jurado et al. [14], whereas data from our [8] senior cohort failed to show any SWS–performance relationships. We should note that these various studies differed methodologically in that Jurado et al. compared SWS of the fastest and slowest thirds of their sample, whereas we used a median split to divide our participants into Fast and Slow Performers in our two studies so as to maximize our statistical power. Nonetheless, the cross-sectional comparisons these studies allow provides some support for the notion [21] that the restorative role of SWS at least for simple neurocognitive functions like reaction time, may diminish across the lifespan. Of course, longitudinal studies would be useful to provide further confirmation for this hypothesis. Admittedly, the current investigation had a number of limitations that warrant consideration. Our sample included normal sleepers and PI sufferers who, by and large, presented to us as research volunteers. Whether our findings apply to middle-aged normals in general and clinical samples of PI patients remains to be determined. It also should be noted that our computer software was calibrated to score a very limited EEG frequency range. As a result, our analyses of these measures connoted differences only in relative power rather than the absolute or total power manifest across the entire EEG spectrum. Given our speculations about endogenous arousal, it may have been useful to obtain power data for those higher frequency bands (e.g., beta, alpha, etc.), which might reflect such arousal. In addition, it should be mentioned that we assessed our subjects’ daytime performance via a limited number of reaction time tasks. As a result, it remains possible that other dimensions of cognitive functioning such as memory, language, reasoning, etc., may show relationships to slow-wave sleep in both the age group studied and in other age groups. Finally, our research approach involving between-subjects comparisons of Fast and Slow performers may have prevented us from discerning the actual importance of SWS to diurnal performance. Perhaps within-subjects comparisons of study participants following both normal/good nights of sleep and nights during with their SWS is experimentally or naturally interrupted would provide somewhat different results than those we obtained. Thus, future studies would benefit by the inclusion of clinical patients, larger research samples, more diverse age groups, a more comprehensive scoring of the EEG spectrum, a wider range of performance measures, and the use of within-subjects methodologies so as to provide a better understanding of importance of SWS to neurocognitive functioning.
134
J.D. Edinger et al. / Physiology & Behavior 70 (2000) 127–134
Acknowledgment This research was supported by the Department of Veterans Affairs Merit Review Program. References [1] American Psychiatric Press. Structured Clinical Interview for DSMIII-R (SCID). Washington, DC: American Psychiatric Press, 1990. [2] Armitage R, Hoffman R, Fitch T. A comparison of period-amplitude and spectral analysis in quantifying sleep EEG: normal controls. Boston, MA: Annual Meeting of the Associated Professional Sleep Societies, 1994. [3] Berry DTR, Webb WB. Sleep and cognitive functions in normal older adults. J Gerontol 1985;40:331. [4] Bliwise DL. Sleep in normal aging and dementia. Sleep 1993;16:40–81. [5] Bonnet MH, Arand DL. 24-hour metabolic rate in insomniacs and matched normal sleepers. Sleep 1995;18:581–8. [6] Bootzin RR. Effects of self-control procedures for insomnia. In Stuart RB, editor. Behavioral Self-Management. New York: Brunner/Mazel, Inc., 1977. [7] Bunnell DE, Bevier W, Horwath SM. Effects of exhaustive exercise on the sleep of men and women. Psychophysiology 1983;20:50–58. [8] Crenshaw MC, Edinger JD. Slow wave sleep and waking cognitive performance among older adults with and without insomnia complaints. Physiol Behav 1999;66:485–92. [9] Edinger JD, Wohlgemuth WK. The significance and management of persistent primary insomnia. Sleep Med Rev 1999;3:101–18. [10] Edinger JD, Morey MC, Sullivan RJ, Higginbotham MB, Marsh GR, Dailey DS, McCall WV. Aerobic fitness, acute exercise and sleep in older men. Sleep 1993;16:351–9. [11] Feinberg I, Koresko RL, Heller N, et al. EEG sleep patterns as a function of normal and pathological aging in man. J Psychiatr Res 1967;5:107. [12] Freedman RR. EEG power spectra in sleep-onset insomnia. Electroencephalogr Clin Neurophysiol 1986;63:408–13.
[13] Honda Y, Takahashi K, Takahashi S, Azumi K, Irie M, Sakuma M, Tsushima T, Schizume K. Growth hormone secretion during nocturnal sleep in normal subjects. J Clin Endocrinol Metab 1969; 29. [14] Jurado JL, Luna Villegas G, Buela–Casal G. Normal human subjects with slow reaction times and larger time estimations after waking have diminished delta sleep. Electroencephalogr Clin Neurophysiol 1989;73:124–8. [15] Letz R, Baker EL. NES2 Neurobehavioral Evaluation System. Chapel Hill: University of North Carolina, 1988. [16] Perlis ML, Giles DE, Mendelson WB, Bootzin RR, Wyatt JK. Psychophysiological insomnia: the behavioural model and a neurocognitive perspective. J Sleep Res 1997;6:179–88. [17] Prinz PN. Sleep patterns in healthy aged: relationship with intellectual function. J Gerentol 1997;32:179. [18] Rechtshaffen A, Kales A. A Manual of Standardized Terminology, Techniques, and Scoring Systems of Sleep Stages of Human Subjects. Los Angeles: UCLA Brain Information Service/Brain Research Institute, 1968. [19] SAS Institute, Inc. SAS Language and Procedures: Usage, ver. 6, 1st ed. Cary, NC: SAS Institute, Inc., 1989. [20] Schramm E, Hohagen F, Grasshoff U, Berger M. Test-retest reliability and validity of the Structured Interview for Sleep Disorders according to DSM-III-R. Am J Psychiatry 1993;150:867–72. [21] Spiegel R, Koberle S, Allen SR. Significance of slow wave sleep: considerations from a clinical viewpoint. Sleep 1986;9:66–79. [22] Spielman AJ, Saskin P, Thorpy MJ. Treatment of chronic insomnia by restriction of time in bed. Sleep 1987;10:45–55. [23] Takahashi Y, Kipnis D, Daughaday W. Growth hormone secretion during sleep. J Clin Investigat 1968;47:2079–90. [24] Tilley AJ, Wilkinson RT. The effects of a restricted sleep regime on the composition of sleep and on performance. Psychophysiology 1984;21:406–12. [25] Vitiello MV, Schwartz RS, Davis MW, Ward RR, Ralph DD, Prinz PN. Sleep quality and increased aerobic fitness in healthy aged men: preliminary findings. J Sleep Res 1992;1(Suppl. 1):245.