Introduction
Most patients in intensive care units (ICUs) experience severe sleep disruptions. These alterations include decreased total sleep time and slow wave sleep, decreased rapid eye movement (REM) sleep and marked fragmentation [
1]-[
6]. They may have clinical consequences in critically ill patients, but few studies have linked sleep disruptions and outcomes [
7].
Quantification of sleep in ICU patients is a major concern since sleep is highly altered and very different from sleep in non-ICU patients, that is, patients with obstructive sleep apnea syndrome. In about one third of ICU patients, the conventional scoring rules of the American Academy of Sleep Medicine (AASM) [
8],[
9] are difficult to use because of altered sleep and wake electroencephalogram (EEG) patterns [
10],[
11] and alternative methods have been recently proposed [
12]-[
14].
When applicable, conventional sleep parameters show high inter-individual variations in ICU populations [
4],[
15]. In addition, patients in ICU display abnormal sleep/wake organization with non-consolidated polyphasic sleep. Patients often spend as much as 50% of their sleep during daytime naps [
1],[
4],[
15], in contrast to the consolidated nocturnal sleep observed in ambulatory patients. Traditional assessments of sleep in this patient population may capture only the gross features of severely disorganized sleep architecture. The relationship between sleep disruptions in ICU and clinical consequences is difficult to establish, prompting a search for additional parameters that could enhance sleep quality assessment.
Based on Bonnet’s sleep continuity theory [
16], which posits that at least 10 minutes of uninterrupted sleep are needed to serve a recuperative function, several authors have deemed quantification of sleep continuity to be of interest. Two studies showed less continuous sleep in patients with sleep-disordered breathing compared to healthy subjects, whereas the usual parameters were not statistically different between groups [
17],[
18]. Regarding the disrupted hypnograms of critically ill patients, we tried to find and develop a simple and pertinent measure of sleep continuity.
Since sleep deprivation has been shown to reduce inspiratory muscle endurance in healthy subjects [
19], we postulated that lack of restorative sleep might impact respiratory functions in ICU patients. Our hypothesis was that sleep episodes lasting less than 10 minutes would be less restorative than sleep episodes lasting more than 10 minutes. We studied sleep in non-sedated ICU patients with hypercapnic respiratory failure treated with non-invasive ventilation (NIV) during several days. We then used NIV outcome as a ‘measure of performance’, that could indirectly investigate sleep’s restorative function.
Our aim was to determine whether or not a sizable proportion of sleep spent in long episodes was associated with NIV success. We also compared sleep continuity during NIV and during spontaneous breathing in a second group and tested the inter-scorer reproducibility of these measures.
Discussion
The primary objective of this study was to search for pertinent quantification of disrupted hypnograms in ICU patients and to perform an advanced evaluation of this new metric. Using a group of patients treated with NIV for respiratory failure, our results demonstrated that patients with NIV success had high sleep continuity, that is, they spent a substantial proportion of their sleep in naps lasting more than 10 minutes rather than in very short sleep bouts. Conversely, patients with NIV failure had low sleep continuity, that is, they spent a high proportion of their sleep in sleep bouts rather than in naps lasing more than 10 minutes. On the other hand, conventional sleep metrics, such as percentage of sleep stages and the Arousal/Awakening Index, did not significantly differ between groups. Our results also showed that NIV had improved sleep continuity by reducing the proportion of time spent in sleep bouts and by increasing the proportion of time spent in long naps. Finally, sleep continuity has a high reproducibility across scorers.
Most patients in ICUs experience severe sleep disruptions including decreased total sleep time and slow wave sleep, decreased REM sleep and marked sleep fragmentation [
1]-[
6]. Sleep fragmentation has captured attention because this index is well-correlated to impaired behavioral performance the day after a fragmented night in healthy subjects [
26],[
27] and because searching for events occurring in the seconds preceding arousal might facilitate the identification of its cause [
5]. However, no studies have demonstrated a significant association between sleep fragmentation and outcome in ICU patients. Our results show that arousal indices cannot distinguish patients with NIV success from those with NIV failure. These findings are in line with the study by Trompeo
et al., which showed a similar fragmentation index in patients with high and low clinical severity scores [
28].
As an alternative to arousal counting, several studies have shown that sleep continuity may provide a pertinent approach allowing quantification of sleep structure and quality [
17],[
18],[
29]-[
33]. To our knowledge, our work is the first study to investigate sleep continuity in critically ill patients. Our results are in line with previous studies in patients showing that measurement of sleep continuity can reveal substantial differences between patients with or without obstructive sleep apnea, differences that are not significant when comparing arousal indexes [
17],[
18]. Our results showed that, during NIV, patients spent less time in bouts and more time in long naps; these findings are congruent with the higher amount of slow wave sleep and REM sleep during NIV. This is in line with the hypothesis that a minimal amount of light sleep continuity is necessary before sleep deepens to N3 and cycles to REM sleep [
30],[
34]. Our findings suggest that, rather than arousal indices, percentage of time spent in sleep episodes lasting less than 10 minutes might be a relevant indicator of sleep alteration in ICU patients. Replication of our findings on a larger group is, nevertheless, required.
Although prolonged PSG, including a large part of the daytime period, is the reference standard for quantifying sleep [
35], several sources of bias may have weakened our results. First, we did not record continuously present environmental stimuli such as noise, light and caretaking activities, and different rates of sleep interruption between the two groups cannot be excluded. Although sleep fragmentation indices were similar in patients with NIV success and NIV failure, it is imaginable that environmental factors had more impact on sleep continuity than on fragmentation. However, these environmental factors account for only 30% of awakenings in ICU patients [
4],[
5]. Moreover, we did not assess circadian rhythms in our patients. Critically ill patients often exhibit a disruption of circadian rhythmicity, and low sleep continuity could be a consequence of circadian disruption and/or misalignment [
36],[
37]. A second caveat is the use of 17-hour recordings, instead of 24-hour PSG with a gap from 9 am to 4 pm. Several studies have emphasized the importance of 24-hour recordings in the ICU [
36],[
37]. Recordings of patients with altered distribution of sleep (that is, more sleep during daytime) could therefore be misinterpreted by the sleep continuity metric. It seems possible that in the hours between 9 am and 4 pm, patients experience sleep/wake patterns differently, depending on their mode of ventilation or degree of disturbance of their circadian time-keeping. Nursing care could also have favored one type of sleep bout and then interfered with our findings. Our results need to be confirmed using 24-hour recordings and a time-of-day effect on sleep continuity needs to be explored. Thirdly, we quantified sleep apnea using thoracic and abdominal inductance plethysmography signals rather than oronasal airflow. Inductance plethysmography has been found reliable in quantification of breath waveforms [
38]. Although the apnea-hypopnea index did not differ between the groups in our study, we cannot rule out the possibility that the patients with lower sleep continuity had severe sleep disruption due to sleep apnea. The fact that sleep continuity is likely to be influenced by a variety of processes ranging from mechanical ventilation and environmental stimuli to the endogenous circadian rhythm adds to its appeal as a physiologic measure of sleep.
Finally, although our results seem promising, we must underline that they originate from a subgroup of ICU patients. The generalization of our results to other ICU patients appears premature. The investigation of sleep continuity measurement on a larger group, including patients with various degrees of severity of illness and various pathologies, needs to be conducted before extended use of our measures can be envisioned.
Our results show that patients with NIV success spent more time in short naps lasting 10 to 30 minutes and less time in sleep bouts. Numerous studies have demonstrated the beneficial effects on behavioral performance of diurnal naps following sleep restriction or sleep deprivation [
21]-[
24]. Our findings are consistent with the sleep continuity theory of Bonnet, which posits that at least 10 minutes of uninterrupted sleep are needed to serve a recuperative function [
16],[
39]. Numerous studies in sleep-deprived healthy subjects have demonstrated that 10- to 20-minute naps exerted a favorable influence with regard to various performance measures [
21]-[
24]. Further work is mandatory to transpose Bonnet’s theory from sleep continuity’s effects on vigilance in healthy volunteers to sleep continuity’s beneficial effects on respiratory performance in ICU patients.
Acknowledgements
The authors deeply thank Karoline Lode-Kolz and Ala Covali-Noroc for reviewing the manuscript, Francoise Zerah and Laurent Margarit for technical support and Jeffrey Arsham for his review and editing of the original English-language manuscript on behalf of CHU de Poitiers. This work was performed at APHP, Hospital Henri Mondor, Creteil, France.
Parts of these results have been presented in abstract form at the APSS annual meeting, June 2013, Baltimore MD, USA.
Competing interests
Dr. Bridoux’s institution received support for travel from Orkyn, Vivisol and Adep. Dr. Brochard has consulted for Drager (SmartCare). His institution received grant support from Drager (SMARTCARE), Covidien (PAV), General Electric (FRC) and Vygon (CPAP). Dr. Drouot has served as a board member for UCB Pharma and received support for travel from Orkyn (ATS meeting 2013). Dr d’Ortho has served as a board member for Bioproject and received support for travel from UCB Pharma. The remaining authors have stated that they have no conflicts of interest.
Authors’ contributions
XD designed the studies, scored the polysomnographies, analyzed the data, compiled the statistics and drafted the manuscript. AB performed the double scoring, analyzed the data of and drafted the manuscript. AWT designed the studies and supervised the recruitment of patients in studies 1 and 2 and drafted the manuscript. FRC recruited the patients of study 1, recorded and analyzed the data and revised the manuscript. ACI recruited the patients of study 2, recorded and analyzed the data and helped to revise the manuscript. SK supervised the statistical analysis, calculated Cohen’s kappa coefficients and ICC analysis and helped to revise the manuscript. LB conceived the studies, participated in their coordination and revised the manuscript. MPO designed the studies and drafted the manuscript. All authors agree to be accountable for all aspects of this work. All authors read and approved the final manuscript.