Effects of Different Sleep Reductions on Daytime Sleepiness

Effects of Different Sleep Reductions on Daytime Sleepiness Alessandra Devoto, Fabio Lucidi, Cristiano Violani, Mario Bertini. Dipartimento di Psicologia, Università degli Studi di Roma "La Sapienza", Roma, Italy. Summary: This study evaluated the effects of different amounts of sleep and SWS restriction on the ensuing day-time sleepiness. Six healthy selected males, after one adaptation night and an initial 8-hr baseline night, were allowed to sleep 5, 4, 3, 2, and 1 hr with a 1-week interval between conditions. The following day, 4 sleep onset MSLT trials and 2 Wilkinson Auditory Vigilance Task (WAVT) were administered. Before each MSLT, self evaluations of sleepiness and activation on a visual analogue scale (ADAS) were assessed. Each restriction night was followed by an 8-hr recovery night, and a final 8-hr baseline night was recorded. The day after each night the same diurnal tests were repeated. Results indicated a linear increase in the propensity to sleep (MSLT) and of subjective sleepiness as a function of the increase in sleep restrictions. Performance scores (WAVT) showed that vigilance is partially affected by sleep restrictions. For each measure, regression analyses showed that the effect of sleep reduction is better predicted by the total duration of sleep than by the amount of SWS. Correlations between measures were negligible with the exception of those between performance and subjective sleepiness measures. Key words: Sleep reduction; SWS loss; daytime sleepiness; MSLT; WAVT: VAS.

CURRENT MODELS of alertness regulation1-4 state that daytime sleepiness mainly depends on the time of day and on the duration of prior waking or sleep. Several studies, manipulating sleep duration, confirmed its relevance on daytime sleepiness intensity: sleep reductions determine a dose-responding increase of sleepiness,5,6 while sleep extensions cause an increase in alertness levels.7,8 However, it has been hypothesised that day-time sleepiness could vary not only as a function of the duration of previous sleep but also as a function of sleep structure. According to Horne9,10 core sleepiness, characterized by the highest sleep pressure, should be due to the deprivation of core sleep, concentrated in the first 3 REM-NREM cycles with the highest amount of Slow Wave Sleep (SWS: stage 3 plus stage 4 of sleep). Whether the SWS amount has a different or more intensive function than the duration of sleep per se on the ensuing day-time sleepiness is an enduring issue. Some studies11-15 addressing this issue by means of the selective deprivation of SWS during the sleep period, showed that sleep

duration is more important than SWS content for daytime alertness. Nevertheless, these studies did not completely clarify the role played by SWS on day-time sleepiness because of some methodological problems mainly due to difficulties in suppressing SWS without modifying sleep duration or continuity. Some indirect evidence for the particular relevance of the amount of SWS on day-time functioning is provided by studies on sleep distribution and duration. Bonnet16 comparing the effects of sleep fragmentation after 1 minute, 10 minutes and 2.5 hours of accumulated sleep, revealed the least performance decreases in the 2.5 hour condition in which the highest amount of SWS was preserved. Lumley, Rohers, Zorick, Lamphere & Roth17 studying the restorative effects of napping for varying durations (0, 15, 30, 60, 120 minutes of sleep), showed that sleepiness decreased until the 60 minute nap; the 120 minute nap with the same amount of SWS and the two-fold duration of sleep did not decrease the level of sleep propensity compared to the 60 minute nap. Carskadon & Dement5,18 evaluated the sleep latencies to the MSLT following one night with sleep restricted to 4 hours and after seven nights with 5 hours of restricted sleep; they found the sleep propensity much more affected with restriction to 4 than 5 hours of sleep. It is worth noting that SWS was not affected when 5 hours of sleep were allowed18 but it was

Accepted for publication March 1999 Comments and Reprint Requests to: Dr. Alessandra Devoto, Ph.D., Dipartimento di Psicologia, Università di Roma "La Sapienza", Via dei Marsi, 78, I-00185, Roma Italy. SLEEP, Vol. 22, No. 3, 1999

336

Daytime sleepiness—Devoto et al

affected when sleep was restricted to 4 hours.19 In 1969, Wilkinson20 performed a study on the effects of various sleep restrictions which could have been crucial to exploring the function of SWS on the daytime alertness. In this study, the sleep duration allowed was 7.7, 5, 3, 2, 1, and 0 hours per night and the ensuing daytime performance was measured by the Wilkinson Auditory Vigilance Task (WAVT). After one night, the results showed a small decrease of performance after sleep restricted to 5 hours and a higher decrease of performance with increasing restrictions (from 3 hours of sleep). Wilkinson suggested that restricting sleep to 5 hours determines no change in the capacity to perform the task because there was no loss of SWS. Only the higher sleep restrictions, determining a loss of SWS, could have impaired the capacity to perform discriminative functions. Unfortunately, this hypothesis could not be verified in that study because of the lack of electropolygraphic recordings. The aim of the present experiment was to compare the effects of SWS loss with respect to sleep time reduction on different measures of daytime sleepiness. For this purpose we compared the effects of finely graded sleep durations of 8, 5, 4, 3, 2, and 1 hours of sleep, electropolygraphically recorded. In the reduction schedule, we postponed sleep onset keeping the final awakening time constant. Considering the prevalent location of the SWS in the first half of the night, this schedule caused a relatively independent decrease in the amount of sleep duration and in the amount of the SWS21 which allowed us to compare the relevance of these two factors on the ensuing daytime sleepiness. The assessment of daytime sleepiness was carried out by using three sets of sensitive measures: (1) the Multiple Sleep Latency Test (MSLT) as an electrophysiological measure of sleep tendency22; (2) an Activation Deactivation Visual Analogue Scale (ADAS)23 as a subjective measure; (3) the Wilkinson Auditory Vigilance Task (WAVT) as a measure of performance impairment.24 The MSLT and the VAS are currently used to provide reliable objective and subjective indices of physiological and subjective sleepiness.25,26 The WAVT is a performance measure introduced since 1969; since then several excellent performance metrics have been developed.27 Nevertheless, the WAVT was employed for two reasons: (1) the present study builds on the Wilkinson classical study (1969) with the further purpose to verify the SWS involvement in performance variation after different sleep restrictions; and (2) in the present partial sleep restriction study we chose a prolonged, monotonous, test-paced discrimination task, like the WAVT, because its test characteristics are classically considered to make sensitive the performance measures of sleep deprivation.28,29 Furthermore, different studies have demonstrated the sensitivity of the WAVT both after total sleep loss24 and after partial sleep restrictions.30 Johnson, Spinweber, Gomez & Matteson,31 in a study SLEEP, Vol. 22, No. 3, 1999

based on between-subject design, found little or no correlation among these measures. Concordance among these measures seems higher when sleepiness is caused by total sleep deprivation32 and lower when determined by partial sleep deprivation.30 The other purpose of the present study was to assess covariations among different measures of sleepiness using a within-subject design in which sleepiness levels were manipulated by partial reductions of sleep time. METHODS Subjects Six healthy male college students (ages 21-25) were paid to take part in the study. Informed consent was obtained from all subjects. They were selected on the basis of their sleep logs, being regular sleepers (7.5-8.5 hr/night) without sleep disorders. Subjects were also selected through an Italian version33 of the MEQ questionnaire34 as intermediate circadian types. All subjects had no pathological score on the Epworth Sleepiness Scale35 and none reported any history of drug abuse, or of medical or psychiatric disorders in a brief interview. Participants were asked to refrain from alcohol, drugs, and caffeine intake throughout the study. Smokers were required to stop smoking 30 minutes before nocturnal and diurnal sessions. In the two nights before every experimental session, subjects were instructed to follow a constant bedtime schedule (00:00 to 08:00 ) actigraphically controlled (by AMI motion logger 16 K). Procedure Each subject followed an experimental schedule in, while after one adaptation night and a first baseline night (BSL1) with 8 hours of sleep (00:00-08:00), sleep duration was restricted each Monday of the following six consecutive weeks. Sleep time allowed in the reduction nights was 5 or 4 or 3 or 2 or 1 hour. Sleep time restrictions were carried out by delaying the sleep onset time and maintaining the awakening time at 08:00 a.m. Each reduction night (RED) was followed, on Tuesday, by a recovery night (REC) with 8 hours of sleep (00:00-08:00). A second baseline (BSL2) of 8 hours was finally performed. In the days following each BSL, RED and REC night four sleep onset MSLT trials (10:00, 12:00, 14:00, and 16:00) and two performance WAVT trials (10:30 and 14:30) were administered. Before each MSLT subjects compiled one Activation Deactivation Visual Scale (ADAS). The baseline nights were located at the beginning and at the end of the conditions to check the potential effects of cumulative sleep pressure due to the restriction schedule. Sleep time restrictions were counterbalanced across subjects dividing them into three blocks: slight restriction (5-4 337


hours of sleep), medium restriction (3 hours of sleep), and strong restrictions (2-1 hours of sleep). Since the Sleep efficiency Index in all the conditions considered, range from a minimum of 93% to a maximum of 99%, the actual Total Sleep Duration (TSD) pairs the Total Time in Bed (TBT). Thus, we will use the two terms in an interchangeable way. Data from the sleep stages within each condition and polygraphic recording techniques have been published elsewhere.21 During sleep time, each subject slept in a darkened single bedroom that was isolated from external time cues and sleep was recorded polygraphically. The nights before the sleep onset time, subjects spent their time in the laboratory, reading, talking, listening to music, or watching TV under the supervision of the experimenters. Dinner was served at 20:00. During the days following every night, subjects were awakened at 08:00, they received their breakfast at 08:30 and they remained in the laboratory to perform the diurnal tests. At 12:45 a light meal was served. Between the diurnal sessions of the tests the participants were allowed to go out of the lab, always in the company of an experimenter. The MSLT followed the standard procedure.36,37 Subjects were instructed at 10.00, 12.00, 14.00 and 16.00 hours to lie down on the bed and try to fall asleep. Standard EEGs (C3-A2 and C4-A1) including an occipital placement (Oz), EMG and horizontal and vertical EOG were recorded during the trials. The sleep latencies (in minutes) were measured considering the time interval between the beginning of the trial (light off) and the first of 3 consecutive 30 seconds epochs of sleep. After sleep onset, subjects were immediately awakened to avoid any cumulative period of sleep. If subjects did not fall asleep, the end of the trial and the sleep latency was established as 20 minutes. The ADAS, given to the subjects before each MSLT trial, is a unipolar visual analogue scale which allows an assessment of how the subject feels "right now" with respect to the following adjectives "sleepy," "energetic," "tired," "calm," "tense." For each adjective subjects respond by making a pen stroke on a 100 millimetre line. Each line anchored at the left end with "not at all" and at the right end with "very much." The distance (in millimetres) of the mark from the left end of the line was the measure of the intensity of the self-evaluation. A computer-controlled version of the WAVT was given in two sessions at 10:30 and 14:30. The WAVT was administered to the subjects by headphone in the soundproof rooms of the lab. Each session lasted 40 minutes on average. During the trials, a series of 500 msec. auditory monotonic signals (background signals) were presented at a rate of two seconds. Randomly interspersed amongst these sounds were "target" signals (with a target signal probability around 0.10) just a little shorter than the "background" signals (c. 400 msec.). Subjects were instructed to respond SLEEP, Vol. 22, No. 3, 1999

to the "target" signals alone, by pressing a console key. Before the beginning of all the sessions, subjects were administered the practice trials to allow familiarization with the task. Furthermore, during the practice an individual level of difficulty of the task was set: the difference between "background" and "target" signals was varied for each subject until performance with 70% of correct responses (hits) was reached. Before every session there was a 1 minute warm-up period. Feedback of the results was given to the subjects once every two trials. For each session, the number of hits and the number of false positives were considered as performance scores. On the basis of the application of the theory of signal detection to a vigilance task, the d prime (d') and the beta (ß) parameters were calculated.38 The d' parameter is a measure of the resolution of the perceptual apparatus and is considered as the intrinsic capacity for a detection of a signal. The ß parameter is an index of the subject's willingness to say that a signal is present; the subject's threshold in answering the signals varies along a "risky-cautious" dimension and could represent the effort with which a subject applies himself/herself to the task.20,39 Data Analysis The measures recorded in the diurnal sessions following the baseline, restriction and recovery nights were considered. In particular, the dependent variables were: 1) the sleep latencies (SL; in minutes) recorded in the four trials of the MSLT; and 2) the scores of the four ADAS administered before each MSLT trial; for this measure the data of the item "sleepy" [ADAS S] and the values of an "activation" index [ADAS A] were considered. The ADAS A index was calculated on the basis of the four scores of the items "energetic" and "sleepy" according to the formula: ADAS A=[(energetic) + 100 - (sleepy)]/2, as suggested by Monk40 3) The scores of the two trials of the WAVT considered the percentage of hits, the percentage of false positives (FP), the d' index, and the ß index. To assess the possible accumulated effects of the sleep restriction schedule, the mean value of each different measure of sleepiness recorded the day following the BSL1 and the BSL2 was preliminary compared using different oneway analyses of variance (ANOVAs). Because no significant differences were found the two baseline day measures were collapsed. The effects of sleep duration on the different sleepiness measures were assessed by two-way repeated measure ANOVAs with SLEEP LENGTH (8 vs. 5 vs. 4 vs. 3 vs. 2 vs. 1 hours of sleep) and TRIALS (4 MSLT, 4 ADAS and 2 WAVT trials). The p values were corrected with the Huynh-Feldt method.41 The WAVT scores were submitted to suitable transformations (i.e., arcosin for the percentages of hits, square root for the percentages of false positives, and logarithmic transformation for the d' and ß indices). All the data were z-transformed to an individual 338


Figure 1. Means of the Total Sleep Duration (TDS) and Slow Wave Sleep (SWS) amounts (in minutes) across the baseline (BSL) and the restriction (RED) nights. 550 500 450 400 350 300

TSD

250

SW S

200 150 100 50 0 BSL1

BSL2

RED5

RED4

basis to reduce between-subject differences in the levels of these variables. To investigate variations in the measures of sleepiness across the different sleep durations, independently of the results of the F omnibus tests, trend analyses considering the linear, quadratic, and cubic components were performed. To evaluate which variable (total sleep duration or slow wave sleep amount) better predicted daytime sleepiness, a multiple regression analysis (method stepwise) was performed on each dependent variable. The variables considered as predictors were: (1) the minutes of total sleep duration (TSD) of the two baseline and five restriction nights; and (2) the amount of slow wave sleep (SWS, minutes of stage 3 NREM plus minutes of stage 4 NREM) in the same nights. The variables considered as criteria were: the mean of the four diurnal trials of the SL, of the ADAS S and of the ADAS A scores and the mean of the two diurnal trials of d' and ß indices and of the FP for the performance measure. These were all collected in the days ensuing the two baseline and five restriction nights. Following Gillberg, Keckelund & Akerstedt,42 we pooled the individual standardised data (with n=7 data points), using the resulting data (with 7X6= 42 data points) for the multiple regressions on each criterion. To evaluate the relationships between the electrophysiological (i.e., SL), subjective (i.e., ADAS S; ADAS A), and performance measures (i.e., d' and ß indices), the correlations (r of Pearson) among the different measures were assessed. As for the previous set of data, the correlations were carried out on the basis of data pooled by standardized SLEEP, Vol. 22, No. 3, 1999

RED3

RED2

RED1

scores of each subject. Each subject, yielded with 28 data points for the electrophysiological and subjective measures, and with 14 data points for the performance indices. The correlation between SL and ADAS S and ADAS A scores were carried out on 6X28=168 data points for each variable. The correlations between SL, ADAS S, ADAS A and WAVT scores were carried out on 6X14=84 data points for each variable. The use of standardized pooled data involves some violations of the assumptions for statistical inference, such as the increase in the degrees of freedom with respect to the actual number of participants. However, in the multiple regressions we were interested in determining which of the two variables predicted each criterion better, and the relationships between predictors and criteria were compared without considering their respective level of probability. To assess the probability level of the correlations between the different measures we corrected the degrees of freedom (df) considering that, for each analysis, one df is lost for each extra subject added to the first in the pooled data set. RESULTS Figure 1 shows the means of Total Sleep Duration (TSD) and Slow Wave Sleep amount (in minutes) across the baseline (BSL) and the restriction (RED) nights. Figure 1 indicates the prevalent location of SWS in the first part of the night, and that the decrease in the amount of TDS and in the amount of SWS exhibits different shapes. 339


sd= 22.7) was different (p=.06) compared to the 12:00 and 14:00 trials (X=55.9, sd= 16.5; X=55.7, sd= 17.9, respectively), but not when compared to the 16:00 trial (X=47.4, sd= 17.6). No other significant effect or interaction was found. Trend analyses were repeated on the main effects of SLEEP LENGTH for each measure. For the SL, results indicated that this effect was due to a linear trend (F1,5=188.30; p=.0000) which explained 81% of the variance. For the ADAS S and ADAS A scores the linear component was significant (F1,5=74.354; p=.0003) and (F1,5=24.467; p=.004), explaining 80% and 87% of the variance, respectively. Results showed the linear component was also significant for d' (F1,510.46; p=.02) and ß (F1,5=12.54; p=.01); this component accounted for 66% of the variance for both WAVT indices. Trend analysis on the FP revealed a quadratic trend (F1,5=9.54; p=.03) which explained 31% of the variance and indicated a sharp increase in the false positives beginning from the reduction with 3 hours of sleep. Figure 2 shows the variations in each measure across the six different sleep durations.

Figure 2. Variations of SL of the MSLT, sleepiness scores (ADAS S) and activation scores (ADAS A), d' index, Beta index and the percent age of false responses (%FP) of the WAVT across different sleep durations. Figures also report the best fitting functions for each measure.

Recovery Sleep on Sleepiness Measures The two-way ANOVAs (SLEEP LENGTH X TRIALS) on each measure after the baseline and the recovery nights (8 hours of sleep) showed a significant main effect for SLEEP LENGTH (F5,25=3.53, p=.01) for the SL. The Tukey test revealed that the SL, after the reduction to 1 hour of sleep, were significantly lower (p