An experimental study of adolescent sleep restriction during a simulated school week: changes in phase, sleep staging, performance and sleepiness

Transcription

1 J Sleep Res. (2017) 26, Adolescent sleep restriction and performance An experimental study of adolescent sleep restriction during a simulated school week: changes in phase, sleep staging, performance and sleepiness ALEX AGOSTINI 1, MARY A. CARSKADON 1,2, JILLIAN DORRIAN 1, SCOTT COUSSENS 1,3 and MICHELLE A. SHORT 1,4 1 Centre for Sleep Research, School of Psychology, Social Work and Social Policy, University of South Australia, Adelaide, Australia; 2 Sleep and Chronobiology Laboratory, EP Bradley Hospital, Department of Psychiatry and Human Behavior, The Warren Alpert Medical School of Brown University, Providence, RI, USA; 3 Cognitive Neuroscience Laboratory, School of Psychology, Social Work and Social Policy, University of South Australia, Magill, Australia; 4 School of Psychology, Flinders University, Adelaide, Australia Keywords adolescence, dim light melatonin onset, performance, sleep loss Correspondence Alex Agostini, Centre for Sleep Research, City East Campus, Playford Building, Level 7, Room P7-3, Frome Road, Adelaide, SA 000, Australia. Tel.: ; fax: ; alex.agostini@unisa.edu.au Accepted in revised form 11 October 2016; received 6 June 2016 DOI: /jsr SUMMARY This laboratory study investigated the impact of restricted sleep during a simulated school week on circadian phase, sleep stages and daytime functioning. Changes were examined across and within days and during a simulated weekend recovery. Participants were 12 healthy secondary school students (six male) aged 1 17 years [mean = 16.1 years, standard deviation (SD) = 0.9]. After 2 nights with 10 h (21:30 07:30 hours), time in bed was restricted to h for nights (02:30 07:30 hours), then returned to 10 h time in bed for 2 nights (21:30 07:30 hours). Saliva was collected in dim light on the first and last sleep restriction nights to measure melatonin onset phase. Sleep was recorded polysomnographically, and the Psychomotor Vigilance Task (PVT) and Karolinska Sleepiness Scale were undertaken 3-hourly while awake. Average phase delay measured by melatonin was 3 h (SD = 0 min). Compared to baseline, sleep during the restriction period contained a smaller percentage of Stages 1 and 2 and rapid eye movement (REM) and a greater percentage of Stage 4. PVT lapses increased significantly during sleep restriction and did not return to baseline levels during recovery. Subjective sleepiness showed a similar pattern during restriction, but returned to baseline levels during recovery. Results suggest that sustained attention in adolescents is affected negatively by sleep restriction, particularly in the early morning, and that a weekend of recovery sleep is insufficient to restore performance. The discrepancy between sleepiness ratings and performance may indicate a lack of perception of this residual impairment. INTRODUCTION Biological and social factors push bedtimes later across adolescence (Carskadon et al., 2004; Wolfson and Carskadon, 1998). Biologically, adolescence is associated with delayed circadian timing system (Carskadon et al., 2004; Crowley et al., 2007) and a reduction in the rate of accumulation of homeostatic pressure throughout the day (Jenni et al., 200; Taylor et al., 200), resulting in a lessened drive for sleep until later in the evening. From a social perspective, reduced parental regulation of bedtimes, greater homework, after-school employment (Carskadon, 1990) and technology in the bedroom (Cain and Gradisar, 2010) may also displace sleep. In addition to the already-present biologically driven delay in circadian timing, delayed bedtimes may contribute further to circadian shifts. For example, Crowley and Carskadon (2010) found that, by delaying bedtime by 1. h for 2 nights, 1- and 16-year-old adolescents had an average phase delay of 4 min (as indicated by dim light melatonin onset; DLMO). The degree to which delayed bedtimes across a typical school week may serve to shift DLMO in adolescents is yet to be investigated. While these factors push bedtimes later, weekday rise times remain constant or become earlier to accommodate 227

2 228 A. Agostini et al. school attendance, resulting in restricted sleep during the school week (Carskadon, 1999). In turn, sleep restriction results in changes to sleep staging, including improved sleep efficiency and a decrease in Stages 1 and 2 and rapid eye movement (REM) sleep, while slow wave sleep (SWS), in particular Stage 4, is maintained (Carskadon and Dement, 1981; Carskadon et al., 1980; Ong et al., 2016). The circumstances producing shorter sleep during adolescence are likely to impair daytime performance and alertness critical during this time particularly for school performance, which may have implications for future success (Roeser et al., 2013; Schmidt and Van der Linden, 201). For example, Beebe et al. (2008) performed an in-home, -night sleep restriction (6. versus 10 h time in bed; TIB) in 20 adolescents (aged years). Results from parent- and adolescent-report measures suggested increased sleepiness and impaired attention, mood and behaviour regulation. In another home-based study, 20 adolescents (aged years) demonstrated slower response times, but no impairment in working memory accuracy or sleepiness during nights of 6 h, compared to nights of 8 h TIB (Jiang et al., 2011). These studies are high in ecological validity but low in experimental control. Indeed, the lack of experimental research in this area has been acknowledged frequently and has been attributed primarily to the logistic and ethical difficulties inherent in working with this population (Beebe et al., 2008). In an innovative approach to addressing these issues, a recent study measured 6 adolescents (aged 1 19 years) in a boarding house during 7 nights of 9 or h TIB, with 2 recovery nights of 9 h TIB (Lo et al., 2016). Results demonstrated that sleep restriction resulted in impairments in average daily sustained attention, working memory, executive function, sleepiness and mood that were not recovered during 2 nights of 9-h opportunities. Sleep in this study was centred (i.e. bedtime was delayed and rise time advanced). Because the sleep pattern for adolescents on school nights often involves sleep restriction with bedtime delayed and rise time constant, it would be beneficial to track performance during restriction with such a sleep schedule. Furthermore, assessing performance daily would be useful to determine whether the impairment accumulated across days, whether performance after nights would be recoverable in 2 nights (i.e. a typical school week, followed by a weekend) and to assess changes in performance across time of day, as this is yet to be conducted in this age group and may reveal important findings for adolescent performance, especially in light of the school start-time debate in the United States (Danner and Phillips, 2008). Therefore, this study examined the impact of experimentally restricted sleep in a sleep laboratory during a simulated school week (i.e. nights of h TIB with delayed bedtimes and constant rise times) on circadian phase (as indicated by DLMO), sleep stages and indicators of daytime functioning (subjective sleepiness and performance), examining changes across and within days, as well as recovery during a simulated weekend (2 nights of 10 h TIB). METHODS Participants Twelve adolescents (six male) aged 1 17 years [mean = 16.1 years, standard deviation (SD) = 0.9] were recruited through newsletters in metropolitan South Australian secondary schools. Inclusion criteria specified that participants: (1) were aged 1 17 years; (2) late- or postpubertal [Tanner Stages 4 or (Tanner, 1990)]; (3) were intermediate chronotype [23 43 on the Composite Morningness/Eveningness Scale (Smith et al., 1989)]; (4) had no medical or psychological disorders and no medication other than the contraceptive pill (assessed by self- and parentreport); and () had no family history of bipolar disorder or epilepsy (assessed by parent-report). Participants were compensated for the inconvenience associated with participation. Ethics approval was granted by the University of South Australia Human Research Ethics Committee. Parents and participants provided written informed consent. Procedure For the days prior to the study, participants were required to go to bed between 21:30 and 22:00 hours each night and rise between 07:00 and 07:30 hours each morning in order to minimize the chances of entering the study carrying a sleep debt. Compliance to this sleep schedule was confirmed by sleep diary and text messages to the researcher immediately before bedtime and after awakening. The study was conducted at the Centre for Sleep Research at University of South Australia. In the sound-attenuated laboratory, temperature was maintained at 21 1 C and lighting at <0 Lux during wake periods. Fig. 1 illustrates the protocol. On the first day and night (10 h TIB), participants became acclimatized to the laboratory and were assessed for sleep disorders (not included in analyses). The following night served as a baseline with 10 h TIB. Participants then had sleep restriction (SR) with h TIB for consecutive nights. The final 2 nights were recovery nights with 10 h TIB. This manipulation was chosen to approximate school week patterns, i.e. shorter sleep during the week and longer sleep at weekends. All sleep periods ended at 07:30 hours to reflect adolescent school night sleep patterns (Jenni et al., 200; Taylor et al., 200). Cognitive test batteries, including the Psychomotor Vigilance Task (PVT) and the Karolinska Sleepiness Scale (KSS) were completed every 3 h during wake, beginning at 08:30 hours. At other times, participants interacted with each other and staff members. Board and card games and a television (to watch DVDs) were provided in the communal lounge room. Participants were monitored constantly to ensure wakefulness. Meals and snacks were provided at set times with no access to food at other times. Free access

3 Sleep and adolescent performance Time P AD Day 1 AD Adaptation 10 h TIB X X X X P X BL Day 2 BL 1 Baseline 10 h TIB X X X *1 X X *2 X Day 3 BL 2 P h TIB X X X X X X Day 4 SR 1 P h TIB X X X X X X Day SR 2 P h TIB X X X X X X Day 6 SR 3 P h TIB X X X *1X X X Day 7 SR 4 P *2 h TIB X X X X P X Rec Day 8 SR Recovery 10 h TIB X X X X P X Rec Day 9 REC 1 Recovery 10 h TIB X X X X X Day 10 REC 2 Practice cognitive test battery *1 First saliva sample X Cognitive test battery *2 Final saliva sample P PSG set up to begin Sleep period Figure 1. Protocol schematic showing temporal placement of cognitive test batteries, sleep and wake periods, polysomnographic (PSG) setups and saliva samples. was provided to water and caffeine-free tea. Caffeine consumption was prohibited during the study. Measures Dim light melatonin onset (DLMO) Saliva samples were taken every 30 min from 17:00 to 22:30 hours on the first SR night and from 17:00 to 02:00 hours on the fifth SR night. Lighting levels during these times were reduced to <30 Lux. Participants were asked to sit up straight and still with their feet flat on the floor for min prior to each sample. Salivettes (Sarstedt AG & Co, N umbrecht, Germany) were used to collect saliva and were frozen at 20 C immediately after collection. Salivary melatonin was analysed using a sensitive (4.3 parts per million: ppm) direct radioimmunoassay with reagents from Buhlmann Laboratories AG (Allschwill, Switzerland) (Voultsios et al., 1997). The intra-assay co-efficient of variation was <10%. The interassay co-efficient of variation was 7.3% at 14.1 ppm and 13.4% at 142 ppm. Values are reported in pg ml 1. DLMO was calculated by linear interpolation across time-points when melatonin concentration increased to 4 pg ml 1 or above. The difference between DLMO following SR relative to baseline was calculated as an indicator of phase shift. Polysomnography Polysomnography (PSG) recordings were conducted for all sleep periods. Electrodes were placed according to the International system (Jasper, 198). The montage included Fp as a general reference and Fp1, Fp2, C3, C4, F3, F4 and O1 referenced to the contralateral mastoid for viewing, left and right electro-oculogram and left and right chin electromyogram. The ground electrode was on the right shoulder. Sleep was subdivided into 30-s epochs and scored according to standard criteria (Rechtschaffen and Kales, 1968) by two experienced sleep technicians (inter-rater reliability = 83%). Variables for analysis included sleep onset latency (SOL; time from lights out until three consecutive epochs of Stage 1 or one epoch of any other stage), total sleep time (TST), sleep efficiency (percentage of TIB spent asleep after lights out), wake after sleep onset (WASO), time in Stages 1 4 and REM. Time in each sleep stage as a proportion of TIB was also calculated in order to represent visually all aspects of TIB in Fig. 2. SWS latency, defined as the first epoch of Stage 3 or 4 sleep after sleep onset, and REM latency, defined as the first epoch of REM sleep after sleep onset, were also analysed. PVT The PVT is a sustained attention task sensitive to the effects of sleep loss (Dorrian et al., 200). Participants completed a 10- min handheld PVT 3-hourly during wake (Fig. 1 shows temporal test placement), with stimuli in the form of a red counter presented at random intervals of 1 10 s. Participants press a button with the thumb of their dominant hand as soon as the counter appears. This stops the counter the number representing their response time in msec. The primary variable in this study was PVT lapses, which for adults are defined as reaction times >00 ms. Research has indicated that this lapse threshold is appropriate for older adolescents (Fallone and Carskadon, 2002), and was therefore used in this study.

4 230 A. Agostini et al. Percentage of TIB BL SR 1 SR 2 SR 3 SR 4 SR REC 1 Night of study REC 2 Stage 1 Stage 2 Stage 3 Stage 4 REM Wake Figure 2. Stacked area graph displaying percentage of time in bed (TIB) spent in wake and each stage of sleep across each night of the study. BL, baseline; SR, sleep restriction; REC, recovery. Dongen et al., 2004). To examine sleep changes, models specified dependent variables of TST, SOL, Stages 1 4, REM and WASO, with a fixed effect of night (1 baseline night, SR nights and 2 recovery nights) and a random effect of subject. Planned comparisons were specified whereby each subsequent night was compared to baseline. For performance and sleepiness, models specified dependent variables of PVT lapses and KSS, with fixed effects of day (1 baseline day, SR days and 2 recovery days), time of day (08:30, 11:30, 14:30, 17:30 and 20:30 hours) and a day 9 time-of-day interaction, with a random effect of subject. Planned comparisons for day of study were specified whereby each day was compared to baseline. Planned comparisons for time of day were specified where each subsequent test was compared to the first test of the day (at 08:30 hours). Degrees of freedom were Satterthwaite-corrected and rounded to the nearest whole number. KSS Participants completed the KSS ( Akerstedt and Gillberg, 1990) 3-hourly immediately following the PVT, rating their sleepiness in the previous min on a nine-point scale with the anchors 1: very alert, 3: alert, : neither alert nor sleepy, 7: sleepy (but not fighting sleep) and 9: very sleepy (fighting sleep). Statistical analyses Mixed-effects models were used to assess changes in sleep, performance and sleepiness within and across days (Van RESULTS DLMO Average DLMO at baseline was at 21:46 hours and following SR was at 00:40 hours, yielding an average phase delay of 3h,SD= 0 min (Table 1). Sleep Sleep variables are summarized in Table 2. TIB was significantly shorter during the SR period compared to both baseline and recovery (P <0.001). Participants obtained Table 1 Participant characteristics Px Age Sex SCME score Habitual SN bedtime Habitual SN wake time Habitual WE bedtime Habitual WE wake time BL DLMO SR DLMO Phase delay (SR-BL) 1 16 M 37 22:30 hours 07:00 hours 23:30 hours 09:00 hours 23:00 hours 01:2 hours 2 h 2 min 2 1 F 33 22:00 hours 06:30 hours 22:30 hours 08:00 hours 22:10 hours 3 16 F 41 23:30 hours 07:30 hours 22:30 hours 09:00 hours * 23:8 hours 4 16 F 30 23:00 hours 07:00 hours 22:00 hours 08:30 hours 20: hours 22:48 hours 1 h 3 min 16 F 38 21:00 hours 06:1 hours 22:00 hours 08:00 hours 22:13 hours 02:30 hours 4 h 16 min 6 17 M 36 22:30 hours 07:00 hours 23:30 hours 09:00 hours 22:10 hours 00:36 hours 2 h 2 min 7 17 M 2 23:00 hours 07:30 hours 24:00 hours 09:30 hours * 02:30 hours 8 1 F 32 21:30 hours 06:30 hours 23:00 hours 09:00 hours 22:1 hours 9 1 F 30 21:20 hours 06:00 hours 22:40 hours 09:30 hours 20:20 hours 23:2 hours 3 h 32 min M 33 22:30 hours 07:30 hours 24:00 hours 10:00 hours 22:14 hours 01:10 hours 2 h 6 min M 37 22:00 hours 07:00 hours 23:30 hours 09:00 hours 21:2 hours 01:27 hours 3 h 3 min M 28 22:00 hours 07:00 hours 24:00 hours 10:00 hours * Age, sex, Smith Composite Morningness/Eveningness (SCME) Scores and habitual school night (SN) and weekend (WE) bedtimes and wake times for each participant. Baseline (BL) and sleep restriction (SR) dim light melatonin onset (DLMO) is also shown, as well as phase delay. M, male; F, female. *Where BL DLMO could not be found, latest sample taken at 22:30 hours. where SR DLMO could not be found, latest sample taken at 02:00 hours. consensus rule applied: if the last sample is below threshold by less than 0% (i.e. melatonin 2 pgml 1 ) the next 30-min point is set as DLMO.

5 Sleep and adolescent performance 231 Table 2 Changes in sleep architecture by night Variable BL SR 1 SR 2 SR 3 SR 4 SR REC 1 REC 2 F TIB (min) *** SD TST (min) *** SD WASO (min) *** SD SE (%) *** SD SOL (min) *** SD Stage 1 (min) *** SD Stage 1 (%) *** SD Stage 2 (min) *** SD Stage 2 (%) *** SD Stage 3 (min) *** SD Stage 3 (%) SD Stage 4 (min) * SD Stage 4 (%) *** SD REM (min) *** SD REM (%) ** SD REM latency (min) SD All subsequent nights are compared to baseline (BL) night. SR, sleep restriction; REC, recovery; TIB, time in bed; TST, total sleep time; WASO, wake after sleep onset; SE, sleep efficiency; SOL, sleep onset latency; SD, standard deviation. Significantly fewer than BL, Significantly greater than BL, ***P < 0.001, **P =0.00, *P =0.01. All df = 7,77. significantly less TST during SR and significantly more sleep during recovery nights compared to baseline (P <0.00). Compared to baseline, SOL was significantly shorter and efficiency significantly higher during SR and recovery (P <0.001). WASO decreased significantly from baseline to SR and returned to baseline levels at recovery (P <0.001). Minutes spent in Stages 1, 2 and 3 and REM sleep decreased from baseline to SR (P <0.001), while Stage 4 sleep increased significantly during SR nights 3 and (P<0.031), and was not significantly different to baseline at recovery. Fig. 2 displays the change in sleep stages across the study, represented as a percentage of TIB, illustrating the reduction in wake and Stage 2 and the increase in Stage 4 during the SR period. PVT lapses PVT lapses across day and time of day are shown in Fig. 3a, b displays mean PVT lapses across days of the study, while Fig. 3c displays mean PVT lapses at each test of the day. There was a significant effect of day (F 7,429 = 13.2, P<0.001) (Fig. 3b), such that there were significantly more lapses during SR and recovery than during baseline (P <0.008). There was a significant effect of time of day (F 4,429 = 4.93, P=0.001), such that there were significantly more lapses at 08:30 hours compared to subsequent tests. The day 9 time-of-day interaction was not significant (F 28,429 = 0.6, P=0.967). However, Fig. 3a suggests that the time-of-day effect may be strongest during the SR period. To illustrate clearly the effect during SR, Fig. 3c displays lapses for each time of day averaged across the SR period only. Subjective sleepiness Sleepiness ratings across day and time of day are shown in Fig. 4a, b shows sleepiness ratings across each day of the study, while Fig. 4c displays mean sleepiness at each test of

6 232 A. Agostini et al (a) PVT Lapses BL SR 1 SR 2 SR 3 SR 4 SR REC 1 REC 2 Day*Test 2 (b) 2 (c) PVT Lapses 1 10 PVT Lapses BL SR 1 SR 2 SR 3 SR 4 SR REC 1 REC Test Day Figure 3. Means and standard deviations of Psychomotor Vigilance Task (PVT) lapses across the study. (a) Results across each testing period of the study; (b) results for each day of the study averaged over the five testing periods; (c) average results at each testing period across the days of the study. The boxes in (a) and (b) are used to highlight the sleep restriction (SR) period. BL, baseline; REC, recovery. the day. There was a significant effect of day (F 7,426 = 4.06, P<0.001) (Fig. 4b), such that ratings were significantly higher during SR (P <0.001) and significantly lower at recovery 2 (P <0.001) compared to baseline. There were significant time-of-day (F 4,426 = 3.0, P=0.008, Fig. 4c) and day 9 time-of-day effects (F 28,426 = 2.40, P<0.001, Fig. 4a). During baseline, subjective sleepiness increased across the day; however, during the SR period ratings are relatively high throughout the day. To highlight this pattern, Fig. 4c displays KSS scores for each time of day averaged across the SR period only. DISCUSSION Results indicate that sleep restricted to h TIB in adolescents may result in phase delays of approximately 3 h across a -night school week. Sleep architecture changes occur, with a reduction in light sleep and REM sleep to accommodate a relative increase in Stage 4. Lapses of attention increase during the restriction period, with worst performance during the morning (08:30 hours). These deficits may not be recovered during a 2-night recovery opportunity, as may be experienced at a weekend. In contrast, the elevated sleepiness observed during restriction reverted to relatively less than baseline levels during the recovery period, potentially indicating a lack of awareness of residual impairment. While results revealed an average phase delay of 3 h after nights, phase delay was not calculable for the whole sample due to missed DLMO for five participants at baseline or sleep restriction. Crowley et al. (2006) found that DLMO (defined as 4 pg ml 1 ) occurs approximately 1. h before habitual bedtime in adolescents aged 1 17 years. We would therefore have expected to capture DLMO in all participants with our protocol. Interestingly, DLMO in this sample was later than expected, given that diaries indicated that participants complied with the previous -day sleep schedule (bedtime: 21:30 22:00 hours, rise time: 07:00 07:30 hours); however, sleep onset latencies were not recorded. Therefore, it may be that sleep onset times were relatively late for this cohort, and saliva sampling not late

7 Sleep and adolescent performance 233 KSS (a) BL SR 1 SR 2 SR 3 SR 4 SR REC 1 REC 2 Day*Test 9 (b) (c) KSS BL SR 1 SR 2 SR 3 SR 4 SR REC 1 REC 2 Day KSS Test Figure 4. Means and standard deviations of the Karolinska Sleepiness Scale (KSS) responses across the study. (a) Results across each testing period of the study; (b) results for each day of the study averaged over the five testing periods; (c) average results at each testing period across the days of the study. The boxes in (a) and (b) are used to highlight the sleep restriction (SR) period. BL, baseline; REC, recovery enough to capture DLMO for everyone. Further, nights of set bedtimes prior to the study may not have been sufficient to phase-advance these participants, as many had relatively late habitual sleep times (as illustrated by Table 1). None the less, the phase delay found in the current study is of similar magnitude to Crowley and Carskadon (2010), which was approximately 23 min day 1 over 2 days with a 1.-h delay in bedtime. Extending this across our -day time-frame would yield an equivalent delay of more than 3 h. This behaviourdriven phase delay may exacerbate the already-present biological phase delay, making sleep difficult to obtain until later in the evening, impacting further the ability to obtain a sufficient amount of sleep before having to wake early for school the next morning. Results demonstrated that sleep restriction alters sleep staging. Compared to longer sleep opportunities, when sleep was restricted, participants obtained less light sleep and REM sleep and more Stage 4 sleep. This is in line with previous studies (Carskadon and Dement, 1981; Carskadon et al., 1981; Ong et al., 2016) and with homeostatic sleep drive theory, given that SWS marked by Stage 4 is a putative marker of sleep drive (Borbely, 1982). Reductions in Stage 2 and REM sleep may impact upon memory consolidation, emotional memory and emotion regulation negatively (Fogel and Smith, 2011; Hutchison and Rathore, 201; Karni et al., 1994). Latency to SWS was shorter during the sleep restriction period compared to baseline, which is also consistent with the theory of homeostatic sleep drive (Borbely, 1982). Our findings also support those of studies in young adolescents, which have found no changes in REM latency between baseline, sleep restriction and recovery (Carskadon et al., 1981). Compared to baseline, recovery nights showed that SOL declined and sleep efficiency and TST were greater, a pattern indicating that the homeostatic drive caused by nights of restricted sleep allowed participants to fall asleep quickly, even in the face of a significant phase delay across restricted nights and with bedtime brought forward by h. Hence, recovery sleep may not be impacted negatively by later bedtimes during the school week. Nevertheless, two recovery nights were insufficient to restore performance to well-rested levels, suggesting that the widely used strategy of using a weekend to recover from the school week may not be fully effective in this population.

8 234 A. Agostini et al. This study extends previous findings by assessing time of day, which has not yet been performed for an adolescent sample. Sleep-restricted participants had more lapses at 08:30 hours compared to later in the day. This finding is consistent with the delay of their DLMO phase and documented circadian rhythm in PVT performance, which peaks in the afternoon, worsening across the night to a nadir at approximately 07:30 hours before starting to improve again (reviewed in Goel et al., 2011). Given that this study was conducted in adolescents who are at an age where they may learn to drive, it is possible that later school start times may result in fewer car crashes (Danner and Phillips, 2008) and better school performance (Edwards, 2012) for those who are sleep restricted. Consistent with Lo et al. (2016), adolescents reported greater sleepiness during nights of sleep restriction than when well rested. In the current study, however, participants reported less sleepiness after 2 recovery nights than during baseline. In contrast, participants in the Lo et al. study reported greater sleepiness during the recovery period than at baseline. The difference in findings may be due to a longer sleep restriction period for participants in Lo and colleagues study (7 nights compared to nights) and a difference in recovery sleep (9 h TIB with an average of 8 h sleep in a boarding house, compared to 10 h TIB with an average of 9. h sleep in a laboratory optimal for sleep). Another difference between the two studies is the sleep timing. While Lo and colleagues centred the restricted sleep timing, rise times were kept constant in the current study, providing a consistent amount of time awake at each performance test across baseline, sleep restriction and recovery. This approach of anchoring rise time is less likely to be ecologically valid for weekends, as adolescents are likely to wake later at weekends rather than going to bed earlier (Carskadon, 1999). Another difference between the two studies is that Lo and colleagues had a control group. Although the within-subjects design reduces the need for a control group, the inclusion of a control group would have strengthened our study. Taken together, our findings suggest that sleep restriction during a school week results in a circadian phase delay and also negatively affects adolescents abilities to sustain attention and maintain alertness. Furthermore, performance is worst in the early morning, which may coincide with a time that adolescents may be driving to school, and thus may represent a safety issue. Finally, the mismatch between performance and subjective sleepiness after weekend recovery sleep could leave adolescents at risk of unknowingly undertaking safety-critical behaviours while impaired. ACKNOWLEDGEMENTS This study was supported by a Rob Pierce Grant-in-Aide from the Australasian Sleep Association and a University of South Australia Divisional Research Performance Fund grant. The authors gratefully acknowledge the assistance of the Adelaide Research Assay Facility, who provided the saliva melatonin assay data and Chelsea Reynolds, Rachel Hadcroft, Melissa Willson, Jordan Watling and Stephen Booth. AUTHOR CONTRIBUTIONS AA contributed to study design, data collection, data entry and analysis and manuscript preparation, MAC to melatonin interpretation and manuscript preparation, JD to data analysis and manuscript preparation, SC to sleep scoring and manuscript preparation and MAS to study design, data collection and manuscript preparation. CONFLICT OF INTEREST The authors report no conflicts of interest. REFERENCES Akerstedt, T. and Gillberg, M. Subjective and objective sleepiness in the active individual. Int. J. Neurosci., 1990, 2: Beebe, D. W., Fallone, G., Godiwala, N. et al. Feasibility and behavioral effects of an at-home multi-night sleep restriction protocol for adolescents. J. Child Psychol. Psychiatry, 2008, 49: Borbely, A. A. A two process model of sleep regulation. Hum. Neurobiol., 1982, 1: Cain, N. and Gradisar, M. Electronic media use and sleep in schoolaged children and adolescents: a review. Sleep Med., 2010, 11: Carskadon, M. A. Patterns of sleep and sleepiness in adolescents. Pediatrician, 1990, 17: 12. Carskadon, M. A. When worlds collide: adolescent need for sleep versus societal demands. Phi Delta Kappan, 1999, 80: Carskadon, M. A. and Dement, W. C. Cumulative effects of sleep restriction on daytime sleepiness. Psychophysiology, 1981, 18: Carskadon, M. A., Harvey, K., Duke, P., Anders, T. F., Litt, I. F. and Dement, W. C. Pubertal changes in daytime sleepiness. Sleep, 1980, 2: Carskadon, M. A., Harvey, K. and Dement, W. C. Acute restriction of nocturnal sleep in children. Percept. Mot. Skills, 1981, 3: Carskadon, M. A., Acebo, C. and Jenni, O. G. Regulation of adolescent sleep: implications for behaviour. Ann. N. Y. Acad. Sci., 2004, 1012: Crowley, S. J. and Carskadon, M. A. Modifications to weekend recovery sleep delay circadian phase in older adolescents. Chronobiol. Int., 2010, 27: Crowley, S. J., Acebo, C., Fallone, G. and Carskadon, M. A. Estimating dim light melatonin onset (DLMO) phase in adolescents using summer or school-year sleep/wake schedules. Sleep, 2006, 29: Crowley, S. J., Acebo, C. and Carskadon, M. A. Sleep, circadian rhythms, and delayed phase in adolescence. Sleep Med., 2007, 8: Danner, F. and Phillips, B. Adolescent sleep, school start times, and teen motor vehicle crashes. J. Clin. Sleep Med., 2008, 4: 33. Dorrian, J., Rogers, N. L. and Dinges, D. F. Psychomotor vigilance performance: a neurocognitive assay sensitive to sleep loss. In: C. Kushida (Ed) Sleep Deprivation: Clinical Issues, Pharmacology and Sleep Loss Effects. Marcel Dekker Inc, New York, 200:

9 Sleep and adolescent performance 23 Edwards, F. Early to rise? The effect of daily start times on academic performance. Econ. Educ. Rev., 2012, 31: Fallone, G. and Carskadon, M. A. PVT performance among children in comparison to adults. Sleep, 2002, 2: 1. Fogel, S. M. and Smith, C. T. The function of the sleep spindle: a physiological index of intelligence and a mechanism for sleepdependent memory consolidation. Neurosci. Biobehav. Rev., 2011, 3: Goel, N., Van Dongen, H. P. A. and Dinges, D. F. Circadian rhythms in sleepiness, alertness and performance. In: M. H. Kryger, T. Roth and W. C. Dement (Eds) Principles and Practices of Sleep Medicine, th edn. Elsevier, St Louis, 2011: Hutchison, I. C. and Rathore, S. The role of REM sleep theta activity in emotional memory. Front. Psychol., 201, 6: Jasper, H. Report of the committee on methods of clinical examination in electroencephalography. Electroencephalogr. Clin. Neurophysiol., 198, 10: Jenni, O., Achermann, P. and Carskadon, M. A. Homeostatic sleep regulation in adolescents. Sleep, 200, 28: Jiang, F., Vandyke, R. D., Zhang, J., Li, F., Gozal, D. and Shen, X. Effect of chronic sleep restriction on sleepiness and working memory in adolescents and young adults. J. Clin. Exp. Neuropsychol., 2011, 33: Karni, A., Tanne, D., Rubenstein, B. S., Askenasy, J. J. and Sagi, D. Dependence on REM sleep of overnight improvement of a perceptual skill. Science, 1994, 26: Lo, J., Ong, J., Leong, R., Gooley, J. and Chee, M. Cognitive performance, sleepiness, and mood in partially sleep deprived adolescents: the Need For Sleep Study. Sleep, 2016, 39: Ong, J. L., Lo, J. C., Gooley, J. J. and Chee, M. W. L. EEG changes across multiple nights of sleep restriction and recovery in adolescents: the Need for Sleep Study. Sleep, 2016, 39: Rechtschaffen, A. and Kales, A. A Manual of Standardized Terminology, Techniques and Scoring System for Sleep Stages of Human Subjects. University of California, Los Angeles, CA, Roeser, K., Schlarb, A. A. and K ubler, A. The Chronotype-Academic Performance Model (CAM): daytime sleepiness and learning motivation link chronotype and school performance in adolescents. Pers. Indiv. Differ., 2013, 4: Schmidt, R. E. and Van der Linden, M. The relations between sleep, personality, behavioral problems, and school performance in adolescents. Sleep. Med. Clin., 201, 10: Smith, C. S., Reilly, C. and Midkiff, K. Evaluation of three circadian rhythm questionnaires with suggestions for an improved measure of morningness. J. Appl. Psychol., 1989, 74: Tanner, J. M. Foetus Into Man: Physical Growth From Conception to Maturity. Harvard University Press, Cambridge, MA, Taylor, D. J., Jenni, O. G., Acebo, C. and Carskadon, M. A. Sleep tendency during extended wakefulness: insights into adolescent sleep regulation and behaviour. J. Sleep Res., 200, 14: Van Dongen, H. P., Olofsen, E., Dinges, D. F. and Maislin, G. Mixedmodel regression analysis and dealing with interindividual differences. Methods Enzymol., 2004, 384: Voultsios, A., Kennaway, D. J. and Dawson, D. Salivary melatonin as a circadian phase marker: validation and comparison to plasma melatonin. J. Biol. Rhythms, 1997, 12: Wolfson, A. R. and Carskadon, M. A. Sleep schedules and daytime functioning in adolescents. Child Dev., 1998, 69: