Christopher B. Miller, Woolcock Institute of Medical Research, University of Sydney, Sydney, Australia/Postal address: PO Box M77, Missenden Road, NSW, 2050, Sydney, Australia. Tel.: +61-(2)-9114-0451; fax: +61-(2)-9114-0014; e-mail: email@example.com
This study profiles changes in self-reported daytime functioning during sleep restriction therapy (SRT) for insomnia. Ecological momentary assessment (EMA) captured point-in-time symptomatology to map the time–course of symptoms. We hypothesized a deterioration (week 1) followed by improvements at week 3 of therapy relative to baseline. Nine patients with psychophysiological insomnia completed the Daytime Insomnia Symptom Scale (DISS) at rise-time, 12:00 hours, 18:00 hours and bedtime for 1 week before and 3 weeks during SRT. Four validated factors from the DISS were analyzed (alert cognition, positive mood, negative mood and sleepiness/fatigue) across 28 days yielding 17 170 data points. Factors evaluated week (baseline versus weeks 1 and 3) and time of day symptomatology. Insomnia Severity Index scores decreased significantly pre-to-post treatment (mean 18 versus 7). Reflecting acute effects of SRT, significant differences were found for all factors, except negative mood, between baseline and week 1 of SRT, suggesting adverse effects. By week 3, sleepiness/fatigue and negative mood decreased significantly compared to baseline, and positive mood showed a trend towards improvement (P =0.06). Sleepiness/fatigue displayed a significant week × time of day interaction, explained by a reduction in sleepiness/fatigue at every daytime assessment point (except bedtime, which remained high). A significant interaction for alert cognition was associated with reduction in alertness at bedtime by week 3 and an increase in alertness at rise-time, suggesting that SRT not only improves sleep, but moderates alertness and sleepiness in therapeutic ways. Initial SRT is associated with an increase in sleepiness/fatigue and a decrease in alert cognition.
Insomnia is a highly prevalent and costly disorder (Morin et al., 2009). It is associated with workplace deficits, increased health risks and health-care utilization (Siebern and Manber, 2010). Further, health-related quality of life has been found to be impaired and it is thought that this drives treatment-seeking behaviour (Kyle et al., 2010).
One of the main effective treatments for insomnia is cognitive–behavioural therapy (CBT-I) (Morin et al., 2006; Smith et al., 2002). Sleep restriction therapy (SRT) has been utilized widely within CBT-I (Edinger and Means, 2005). Indeed, it is almost always included in single or brief interventions to treat insomnia (Buysse et al., 2011; Edinger and Means, 2005). Anecdotally, it is believed to be one of the most active components within CBT-I. Despite this, our current understanding of how and why this potent and difficult behavioural intervention may work is poorly understood (Morgenthaler et al., 2006). Further, adherence and the perceived benefits of this therapy have been linked previously to a positive treatment response (Harvey et al., 2002); however, it may also cause some adverse effects (Kyle et al., 2011). In order to understand further the daytime impact, mechanisms and adverse effects of this brief behavioural therapy, novel measurement approaches must be implemented.
Ecological momentary assessment is a methodology which allows participants to report aspects of behaviour prospectively and repeatedly in many conditions (Shiffman et al., 2008). It permits reporting of ecologically valid symptoms within the context of their own environment (Buysse et al., 2007), and so may be useful to profile the daytime impact, potential mechanisms of action and adverse effects of stand-alone SRT on a day-by-day basis through the course of treatment.
The aim of this study was to profile self-report daytime functioning over the course of 3 weeks of SRT for insomnia. We hypothesized that daytime levels of alert cognition, sleepiness/fatigue, positive and negative mood from the Daytime Insomnia Symptom Scale (DISS; Buysse et al., 2007) would deteriorate initially in line with an acute restriction of sleep opportunity, but then improve beyond baseline levels through SRT in line with a treatment response.
Nine participants who responded to advertisements were recruited from the general population and were screened for psychophysiological insomnia (Edinger et al., 2004). Participants slept for 2 nights in the University of Glasgow Sleep Centre, Scotland, prior to study enrolment as part of a concurrent insomnia trial (grant no. R01MH077901). Each subject gave informed consent to participate in the study.
Two nights of polysomnography (PSG) were used to exclude sleep-related comorbidities and to acquire a baseline assessment of sleep. A standard PSG montage was used, involving electroencephalographic [EEG: Fp1 (neutral), C3, P3 (reference), O1, Fpz, Fz, Cz, Pz, Oz, F4, C4], electro-oculographic (EOG: horizontal and vertical) and electromyographic (submental) recordings. On night 1, all participants were screened for sleep-disordered breathing and periodic limb movements through monitoring of abdominal and thoracic effort, nasal airflow, oximetry and bilateral tibialis anterior EMG. Sleep was recorded on a Lifelines Trackit ambulatory recorder (Over Wallop, England, UK) and scored visually by two experienced scorers (>90% interscorer reliability) according to criteria by Rechtschaffen and Kales (1968). For study inclusion, patients were required to have an apnea hypopnea index (AHI) and periodic limb movements of sleep (PLMS) index < 10. This study was reviewed and approved by the West of Scotland NHS research ethics committee (protocol no. 10/SO701/85).
SRT comprised one 40-min face-to-face session, using Spielman et al.'s (1987) instructions, with minimum time in bed set at 5 h. The intervention was standardized and supported by a brief set of digital slides (eight in total) and an information manual (11 pages, in plain English). Two further in-person sessions and two telephone calls (each 5–10 min) were provided to review sleep efficiency and titrate the sleep window. For titration, time in bed was modified on a weekly basis. Changes were made to time in bed incrementally, depending on the achievement of ‘good sleep efficiency’: 85–89% no change, ≥90% increase by 15 min, less than 85% decrease by 15 min (cf. Spielman et al., 1987). Time out of bed was rarely changed. In nearly all cases time to bed was modified in line with Spielman et al.'s (1987) initial instructions (see Table 1). No other CBT-I components were addressed. The experimenters (S.D.K., C.B.M.) acted as therapists in this study. Home adherence to SRT was monitored via a subjective sleep diary (throughout therapy), and also via a specific sleep restriction adherence scale (SRAS; Kyle and Crawford, unpublished), completed on a week-by-week basis. The SRAS is based loosely on the Medical Outcomes Study general adherence scale (MOS-A; Kravitz et al., 1993), but is modified to probe adherence to different aspects of SRT using five specific questions (e.g. item 4: ‘I got up at my calculated “rising time” on weekdays…’/response selection: 1 = ‘none of the time’; 2 = ‘a little of the time’; 3 = ‘some of the time’; 4 = ‘a good bit of the time’; 5 = ‘most of the time’; 6 = ‘all the time’). Possible total SRAS scores range from 5 to 30, with higher scores being indicative of greater levels of adherence. Preliminary (unpublished) psychometric evaluation of the SRAS, with 42 insomnia patients undergoing SRT, revealed high levels of internal consistency (Cronbach's α = 0.92; range of item-deletion α = 0.89–0.93, mean α = 0.91). The Insomnia Severity Index (ISI) (Morin, 1993) was completed before and after therapy to monitor more global effects of treatment. Pairwise comparisons were used to evaluate differences.
Table 1. Mean sleep diary data, Sleep Restriction Adherence Scale (SRAS) and Daytime Insomnia Symptom Scale (DISS) factor scores at baseline and through sleep restriction therapy
Mean and SD
Mean and 95% CI
Prescribed time in bed (apart from baseline)
SE, standard error; SRT, sleep restriction therapy; TIB, time in bed; TST, total sleep time.
The top half of the table describes the mean sleep diary data with standard deviations (minutes) for total sleep time and prescribed time in bed (weeks 1–4 of SRT), the mean and standard deviations (SD) of Sleep Restriction Adherence Scale (SRAS) scores (for weeks 1–4 of SRT) and the mean [with 95% confidence interval (CI)] for the four daytime symptoms variables over the 4 weeks of observation. The bottom half of the table tests for differences in total sleep time, time in bed and SRAS results. This also tests our specific hypotheses about a transient worsening of mood in response to SRT at week 1 followed by an overall improvement over the course of treatment by week 3.
Participants completed paper-based versions of the DISS at four assessment points per day (rise-time, 12:00 hours, 18:00 hours and bedtime) for 1 week before the SRT intervention (baseline) and for 3 weeks during the intervention (weeks 1, 2 and 3), representing the acute treatment phase (see Fig. 1). The DISS consists of 20 visual analogue scales (ranging from 0 to 100 on a 100-mm line. The participant is asked to mark along the line where they agree or disagree with each of the 20 statements at the specified moment in time. For example, question 1 asks: ‘How alert do you feel’, with ‘very little’ at the extreme left of the 100-mm line and ‘very much’ at the right.
Eighteen of the 20 items were utilized, forming four previously validated factors; sleepiness/fatigue (containing the adjectives sleepy, fatigued and exhausted), negative mood (anxious, stressed, tense, sad and irritable), positive mood (relaxed, energetic, calm, happy and efficient) and alert cognition (forgetful, clear-headed, concentrate, effort and alert) (see Buysse et al., 2007). In order to improve adherence to this procedure, participants were offered SMS (short message service) text message reminders to their mobile phone. Two participants opted to receive these at the 12:00 hours and 18:00 hours time-points only.
A linear mixed-model analysis using SPSS software (SPSS IBM version 19.0.0; IBM, Armonk, NY, USA) was implemented for each factor. Fixed effects included week (baseline, 1, 2, 3) of the intervention and for time of day of DISS completion (rise-time, 12:00 hours, 18:00 hours and bedtime). Random effects were run to account for between-subject variation. Only random effects were included for the intercepts. Uncorrected pairwise comparisons based on estimated marginal means were implemented to investigate the primary outcomes.
In total, nine patients (six female; mean age 46.4 years, range 34–58) screened thoroughly for psychophysiological insomnia (Edinger et al., 2004) completed this study. No participants were taking prescribed sleep-promoting hypnotics during the course of this study.
Confirming the expected therapeutic benefits of SRT, sleep diary data revealed significant and robust improvements, baseline to week 4, in both sleep efficiency (73 versus 93%; P <0.01) and sleep quality ratings (1.8 versus 2.3; P <0.01). Significant pre-to-post reductions (P <0.001) were found for: sleep onset latency (33.7 versus 10.7 min), time awake after sleep onset (58.7 versus 15 min), number of awakenings during the night (2.5 versus 1.6) and total time spent in bed (489 versus 367 min). Total sleep time reduced marginally (357 versus 340 min); this was not significant (P >0.05). A significant reduction was found for both total sleep time (357 versus 285 min; P <0.001) and time in bed (489 versus 355 min; P <0.001) between baseline and week 1 of SRT (see Table 1). ISI scores decreased significantly at post-treatment [mean = 18 (5) versus 7 (5), P <0.05]. The SRT self-report adherence mean scores remained high across the weeks of therapy [week 1 = 26.1 (3.5), week 2 = 24.0 (3.8), week 3 = 25.1 (1.3), week 4 = 23.1 (6.1)], with a range of 10–30.
Turning to the primary purpose of the study, the completion rate for the DISS questionnaire throughout the 4 weeks of data collection was highly satisfactory at 94.6% (ranging from 76 to 100%), yielding 19 075 data points of a possible 20 160. Each participant was asked to complete the 20-item DISS four times per day every day for 4 weeks (baseline, 1, 2, 3). For the primary analysis, 18 questions evaluated four dimensions of interest: sleepiness/fatigue, alert cognition, positive and negative mood. This resulted in 17 170 data points (of a possible 18 144). The findings are reported in the following sections.
In order to first determine differences for time of questionnaire completion, we evaluated potential daytime patterns to each of the four factors by analysing the main effect of time of day. Significant differences were found across the four daytime assessment points for three factors: sleepiness/fatigue, alert cognition, positive mood (F(3, 930), sleepiness/fatigue = 63.97, alert cognition = 82.93, positive mood = 42.28; all P <0.001). Negative mood displayed no diurnal change (F(3, 930) = 0.48; P =0.699).
To test the initial hypothesis for a deterioration in symptoms at week 1 of SRT, main effects were analysed for week, yielding significant results for all four factors (F(3, 930), sleepiness/fatigue = 14.30, alert cognition = 6.54, negative mood = 7.75, positive mood = 6.81; all P <0.001). We then compared baseline to week 1 of SRT for all factors. Pairwise comparisons revealed a significant increase for sleepiness/fatigue (P <0.01), a significant decrease for alert cognition (P <0.001) and positive mood (P <0.05) and no change for negative mood. This deterioration in symptoms at week 1 was in line with our initial hypothesis regarding early adverse effects due to acute sleep restriction as part of therapy. To test the second part of our hypothesis (that daytime insomnia symptoms would reduce through SRT), we evaluated differences between baseline and week 3. Consistent with expectation, a significant decrease (P ≤0.01) was found for sleepiness/fatigue and negative mood, with a trend for improvement in positive mood (P =0.06). Alert cognition, however, was not significantly different (see Table 1).
Next, a time of day × week interaction was used to examine potential diurnal changes in the daytime assessment of symptoms through SRT between baseline and week 3. Sleepiness/fatigue (F(9, 930) = 2.54; P <0.01) and alert cognition (F(9, 930) = 2.42; P <0.05) displayed significant interactions. This was non-significant for positive (F(9, 930) = 1.24; P =0.266) and negative mood (F(9, 930) = 0.36; P =0.956). Specifically, pairwise comparisons revealed a decrease at week 3 compared to baseline for sleepiness/fatigue at rise-time [61.9, 95% confidence interval (CI): 51.6, 72.2 versus 52.2, 95% CI: 42.0, 62.5; P <0.01), 12:00 hours (47.9, 95% CI: 37.6, 58.2 versus 37. 0, 95% CI: 26.7, 47.3; P <0.01) and 18:00 hours (48.3, CI: 38.1, 58.6 versus 39.6, CI: 29.3, 49.9; P <0.05) time-points. The bedtime time-point increased, but not significantly (P =0.172). For alert cognition, comparisons revealed a significant increase for rise-time (42.2, CI: 33.8, 50.5 versus 48.5, CI: 40.2, 56.7; P <0.05) and a significant decrease at the bedtime assessment point (49.4, CI: 41.1, 57.712 versus 40.3, CI: 32.0, 48.6; P <0.01) at week 3 of SRT, with no significant differences for the 12:00 hours and 18:00 hours time-points (see Fig. 2).
As a prelude to discussing the impact of SRT on daytime variables, it is important to first note the following about this exploratory treatment study: participants adhered to the SRT instructions (as evaluated by the SRT adherence scale), SRT was effective in improving sleep through a reduction in insomnia severity and the completion rate for the DISS questionnaire was high.
Our primary aim was to examine self-report daytime functioning during the acute course of SRT and to evaluate any initial impairment compared to baseline. Initial implementation of SRT led to significant elevations in sleepiness/fatigue and significant reductions in alert cognition and positive mood (week 1 compared to baseline), suggesting a general deterioration in symptoms in line with previously observed adverse effect data (Kyle et al., 2011). Negative mood remained unchanged. Our data suggest a reduction of symptoms by week 3 of SRT compared to baseline, with sleepiness/fatigue and negative mood both reducing significantly. Specifically, sleepiness/fatigue reduced at rise-time, 12:00 hours and 18:00 hours, but the bedtime assessment increased (although not significantly). Alert cognition did not change overall but, instead, showed changes at specific time-points, increasing at rise-time and decreasing at bedtime (both significantly), suggesting improvements due to SRT (see Fig. 2).
Modification in daytime ‘alert cognition’ may suggest a potential mechanism by which SRT exerts its therapeutic effect. That is, SRT may help to restructure cognitive arousal, inhibiting it at sleep onset (to permit sleep) and restoring it at rise-time (to permit optimum daytime functioning). Interestingly, baseline alert cognition scores resemble those found previously in insomnia; by week 3 of treatment, however, these scores resemble those of good sleeping controls (cf. Buysse et al., 2007). Specific testing of this normalization of alert cognition at rise-time and bedtime may help to uncover the role of cognitive arousal in insomnia in response to effective behavioural treatment. Further trials are required with untreated waiting-list controls in order to understand fully if these changes are related to SRT.
Further, the increase in the factor sleepiness/fatigue at bedtime at week 3 of SRT is indicative of heightened sleep pressure (also reflected in reduced sleep latencies). This may help to overcome cognitive arousal (as described previously in the reduction in alert cognition), suggesting validation of alterations to the sleep experience and cognitive arousal facilitating the initiation of sleep. This has been postulated previously as a mechanism of action within SRT (Kyle et al., 2011; Pigeon and Perlis, 2006; Spielman et al., 1987). Future studies, with larger samples and in the context of a more controlled design, should assess whether changes in sleepiness/fatigue relate to changes in cognitive arousal. As for the two affect factors, positive mood increased overall through SRT (although not significantly), with a slight trend for higher scores at rise-time. Negative mood displayed no daytime differences (unlike the other three factors). This may be attributable to these factors all assessing the same fundamental construct, a lack of power to test this hypothesis or no dysphoric symptoms within the subject group. Of note, negative mood reduced significantly at week 3 compared to baseline (the mean difference was only 3.6; see Fig. 2).
This is the first study to map changes in daytime functioning, using an EMA micro as opposed to a macro between-subjects approach using brief SRT for psychophysiological insomnia. Previously, researchers have focused upon differences between groups of insomniacs and good sleepers in order to profile differences in daytime functioning (Buysse et al., 2007; Levitt et al., 2004). Two studies have validated the use of the DISS via EMA methods in insomnia and healthy good sleepers, but not through treatment. However, it is difficult to compare the present study findings with those of previous studies, as these initially used a non-validated version of the DISS (Levitt et al., 2004) and also solely reported factor scores for the groups of insomniacs and good sleeping controls (Buysse et al., 2007). One other previous study has attempted to compare the differences between good-sleeping control participants and insomniacs via ambulatory assessment (Varkevisser et al., 2007). In this study, the authors attempted to probe differences in performance and wellbeing (concentration, fatigue, mood and sleepiness). No differences were found regarding the performance measures, but they discovered that subjective wellbeing was compromised in insomniacs compared to controls. Further trials using EMA methods are now required to evaluate the role of daytime symptoms in response to other single components of CBT for insomnia.
It is also important to note that although the term ‘sleep restriction’ (SR) is used more widely in sleep science—usually in studies where healthy participants are systematically denied a predefined period of sleep to investigate the effects of sleep loss upon cognitive and physiological functioning (Dinges et al., 1997; Van Dongen et al., 2003)—the context between these protocols and the present study, where SRT is used therapeutically, differs greatly. Nevertheless, some of the effects of SR in these carefully conducted laboratory studies are what have prompted our taking a careful look at the potential adverse effects of SRT insomnia patients. The results suggest that total sleep time reduced significantly between baseline and week 1, but not between baseline and week 4 of SRT. Therefore, the changes that we see at week 1 in the three factors (alert cognition, sleepiness/fatigue and positive mood) may be attributable to changes in perceived sleep time. It is possible that participants adapted over time to the initial effects of this type of sleep restriction by week 3 (Belenky et al., 2003). However, this does not explain fully the significant reduction in sleepiness/fatigue and negative mood at week 3 compared to baseline levels. Further controlled studies are required to evaluate acute sleep restriction in patients with insomnia compared to healthy good-sleeping controls.
A number of limitations regarding this study must be addressed. A regression to the mean may explain the reduction in daytime impairments in alert cognition and sleepiness/fatigue. However, an initial decrease in alert cognition and an elevation in sleepiness/fatigue at week 1 and the subsequent reversal of this at week 3 suggests a contrary conclusion. The experimenters (S.D.K., C.B.M.) also acted as therapists in this study; therefore, it is difficult to rule out experimenter, performance and demand effects. It could be argued that a lack of participants in this study represents low statistical power and increases the probability of obtaining a type II statistical error. Also, the precision of estimating unknown parameters may be compromised by the lack of statistical power. Participants were asked to respond repeatedly to items on a visual analogue scale, a fine-grained continuous outcome measure with increased statistical power compared to more standard ordinal measures. Further, we were not able to determine objective adherence to EMA questionnaire completion. Previously, hand-held computers with time-stamped recordings have been used to determine exactly when a participant completes a questionnaire (Buysse et al., 2007). The use of such devices may be more reliable, but they also bring about difficulties such as battery life and learning how to use them.
In summary, results demonstrate that the DISS, completed via EMA, can profile changes in self-report daytime functioning through SRT for psychophysiological insomnia. Changes at rise-time and bedtime for alert cognition and sleepiness/fatigue seem to be candidates for further study as mechanisms of action in treatment response, potentially reflecting changes to the input and output of the sleep homeostat and cognitive arousal. Further objective assessments of insomnia treatment via repeated measures are required in order to elucidate the role of sleep and arousal in response to SRT for insomnia. Results may help to uncover mechanisms of action within single components of CBT-I. Those delivering SRT as part of CBT-I should be mindful of an initial deterioration in symptoms at week 1 (cf. Kyle et al., 2011). It may be necessary to tailor SRT to suit the needs of the patient in order to reassure and overcome the initial negative symptoms of therapy. Such understanding may help to disseminate this treatment option safely.
This research was supported by the Chief Scientist Office (Scotland), grant no. CZG/2/503; the National Institutes of Health (NIH) grant no. R01MH077901; the National Health and Medical Research Council (NHMRC, Australia) Centre for Integrated Research and Understanding of Sleep (CIRUS), grant no. 571421; and the Sackler Institute of Psychobiological Research (University of Glasgow Sleep Centre).