The main findings from our study highlight a progressive decrease in PSI along with sleep depth based on PSG data. There was also a significant correlation of PSI with the ratio of the stage N2 and RCSQ for nighttime sleep. Sleep assessment was easily conducted using SedLine, with the exception of several artifacts in a single case. Importantly, the frequency of patients’ refusal or complaints was much higher with PSG than with SedLine.
Our assessment of sleep architecture in the 24-hour PSG data described an increase in light sleep and a reduction of restorative sleep in critically ill patients, despite preserved quantity of nighttime sleep. Our findings match those previously reported [<link rid="bib1">1</link>–<link rid="bib4">4</link>]. Anderson et al. showed that the sleep architecture among those in the critical care settings included 11–59% of stage N1, 26–74% of stage N2, 0.15-22% of stage N3, and 1–12% of REM . Widely varying ranges among patients, as well as among studies, are likely to reflect the heterogeneity of this clinical population in respect to differences in diseases, disease severity, or medications [<link rid="bib5">5</link>]. Although most patients in our small study population were postoperative cardiovascular patients and their sleep measurements were conducted appropriately on the 4th or 5th day after ICU admission, the PSG results suggest high interindividual differences between conditions.
PSI decreased along with sleep depth, but its distribution was not equal between sleep stages. The PSI values between WK and stage N1 were quite close compared to those from stage N3 and REM, whereas the wide distribution of stage N2 largely overlapped with the PSI values of stage N1 and REM. A previous study measuring sleep depth in healthy subjects with BIS produced similar results, but values of REM presented distributions closer to the ones observed during stage N1 [<link rid="bib17">17</link>]. The overlap between consecutive sleep stages could also be explained by the continuity of the sleep process, with soft transitions between sleep stages [<link rid="bib17">17</link>]. Vacas et al. reported that SedLine seemed to identify WK or stage N2 when PSG would label the same as stage N1 [<link rid="bib16">16</link>]. Further, atypical PSG signs in critically ill patients, such as a lack of K-complex or spindles that are typically indicative of stage N2, could not be distinguished by SedLine[<link rid="bib4">4</link>].
RCSQ values describing sleep satisfaction also showed wide variance among subjects in the present study. Similar to the PSG results, this may reflect interindividual differences from diverse medical conditions. Importantly, RCSQ showed a significant positive correlation with the nighttime ratio of stage N2 alone or combined with stage N3 and REM as well as a negative correlation with the nighttime average PSI. A significant correlation was also found for one of the five RCSQ subscales assessing sleep depth. Subjective sleep evaluation is considered to depend on sleep depth, as suggested by Richards et al . In the present study, RCSQ was found to correlate with stage N2 alone or combined with N3 and REM. Since most of the sleep time recorded in our study consisted of epochs scored as stage N1 or N2 and restorative sleep was strongly reduced, RCSQ results suggest that satisfactory sleep could be achieved with adequate time in stage N2.
In previous studies comparing BIS and PSG, stage N1 and N2 were combined in a broad category of light sleep [<link rid="bib17">17</link>]. However, we separated stages N1 and N2 because subjective feelings of good sleep were positively correlated with time spent in stage N2 or with N3, REM, or both N3 and REM in our study. This is the main reason why we analyzed and provide the ROC of PSI in Fig. 3, indicating the threshold necessary to distinguish between stage N1 and N2 as well as the respective high specificity, sensitivity, and AUC. The cutoff value of PSI can be interpreted as a target for sleep depth to fulfill subjective sleep quality rather than detecting sleep or awake states. The specificity may not have been as high as the sensitivity due to the overlap between consecutive sleep stages and the potential difficulty in identifying stage N2 in the critical population, as already mentioned. Nevertheless, the precise quantity requirements of each sleep stage could not be identified in our study.
Some study limitations should be acknowledged. First, our sample size was small and resulted from recruitment at a single treatment center. This restricts our ability to generalize our results. Although more than 400,000 components of data were analyzed to validate PSI against PSG, the correlation between objective and subjective sleep evaluation was based on the data of only 10 subjects. We should also consider the interacting influence of the processed EEG monitors in the ICU that are different from those typically applied during general anesthesia . As the heterogeneity among the 10 study participants also limits the data, future multi-center studies with larger sample sizes and a variety of critical care patients may provide a clearer picture. Importantly, the sleep measurement devices should be validated in healthy subjects as well.
Second, the accuracy of the PSG scoring was not verified in this study. The small study population included three delirious patients, and several patients were treated with small doses of fentanyl or dexmedetomidine, with or without mechanical ventilation. It has been previously reported that restorative sleep is reduced in delirious patients or opiate users [<link rid="bib21">21</link>]. Moreover, an increase in stage N3 with slow waves is commonly observed in cases of encephalopathy [6, 22, 23, 24]. Dexmedetomidine could then improve SE, increasing the amounts of stage N2 relative to stage N1, but without increasing restorative sleep [<link rid="bib25">25</link>]. Regardless of delirium, mechanical ventilation, or sepsis, various atypical PSG findings, including the absence of particular waveforms in stage N2 or the discrepancy between behavioral assessment and PSG staging (so-called “pathological wakefulness”), are detectable in critically ill patients [<link rid="bib6">6</link>, <link rid="bib23">23</link>, <link rid="bib24">24</link>]. However, when, why, or how often these may happen remains unknown. Additionally, due to these abnormalities, the appropriateness of PSG scoring using the standard classification is suspicious in the context of the ICU[<link rid="bib4">4</link>, <link rid="bib6">6</link>, <link rid="bib7">7</link>, <link rid="bib24">24</link>, <link rid="bib26">26</link>]. A modified classification has been proposed [<link rid="bib4">4</link>, <link rid="bib7">7</link>, <link rid="bib22">22</link>, <link rid="bib24">24</link>, <link rid="bib26">26</link>], but has not yet been validated. The same concerns may apply to PSI, since it is converted from EEG signals that are used to evaluate the depth of sedation. For this reason, comparing objective and subjective sleep metrics as well as verifying their association is necessary.
The third limitation of this study is that factors potentially affecting subjective sleep evaluation could not be precisely identified, although our results suggest that a sleep depth of stage N2 or deeper is required for sleep satisfaction. Importantly, we found PSI to be related to sleep depth and to allow the discrimination between sleep depth degrees with high sensitivity and specificity. However, not only sleep depth was identified to be important for subjective sleep quality, but also sleep quantity at each depth. Yet, the length of deeper sleep that is necessary to achieve satisfactory sleep quality is still unclear. Frequent sleep fragmentation, commonly observed in critical care settings, was unremarkable in our results. The effects of sleep continuity on sleep quality as well as the required quantity of deep sleep should be further investigated.
Finally, stage N2 sleep produced subjective sleep satisfaction, but this alone may not be sufficient for truly good sleep. The most important domain of sleep to maintain physical or psychological well-being is the restorative sleep obtained during stage N3 and REM, which was highly impaired in our patients. Thus, further validation should be conducted in healthy populations with normal sleep architectures. Further investigation with a larger cohort of patients is needed.