What this is
- This research investigates the connections between sleep quality, circadian rhythms, and dynamics.
- It introduces a new EEG-derived indicator based on timing.
- The study utilizes data from wearable EEG devices and polysomnography to analyze sleep patterns across different age groups.
Essence
- The timing of correlates with circadian phase indicators and subjective sleep quality. A new REM-derived metric shows potential as a indicator.
Key takeaways
- timing (REM) correlates with subjective and actigraphy-derived sleep midpoints. This suggests that REM can serve as a reliable indicator of circadian phase.
- Children and middle-aged adults exhibit earlier REM timing compared to teenagers and young adults, reflecting age-related changes in circadian rhythms.
- No significant association was found between variables and subjective sleep quality measures, indicating a complex relationship that requires further investigation.
Caveats
- The study's reliance on EEG headbands instead of traditional polysomnography limits the ability to perform comprehensive circadian analyses.
- Exclusion of extreme prevents generalization of findings to all populations, particularly those with atypical sleep patterns.
- The lack of longitudinal data restricts the ability to assess interindividual differences in sleep architecture over time.
Definitions
- Chronotype: Individual-specific timing of sleep-wake patterns influenced by circadian rhythms.
- REM sleep: A sleep phase characterized by rapid eye movement, associated with dreaming and memory processing.
AI simplified
Introduction
Efficient, satisfactory sleep is a complex concept that is difficult to characterize along a single dimension or using only one type of metric. Besides focusing on objective PSG measures, such as duration, continuity, or composition, the proper timing of sleep driven by the circadian component of sleep regulation is also an important factor. Optimal timing relies on the internally generated circadian rhythm showing individual-specific synchronization to, and phase alignment with, rhythmic environmental ques (Earth’s rotation around its own axis and consequent day–night alternation). The interindividual differences in the phase of entrainment of the internal and environmental rhythms are treated within the context of the chronotype concept. The latter is frequently assessed by means of questionnaires targeting habitual sleep–wake schedules (Roenneberg et al., 2022). Although there are several available objective methods for determining circadian phase non-invasively such as mobile core body temperature measurement or actigraphy recordings, the estimation of chronotype is not a usual part of routine diagnostic or research-related examinations of sleep. This represents a significant gap in the field not only because participants with different chronotypes are in different circadian phase during the same period of sleep recording (Bódizs et al., 2022) but also because of the known association of chronotype with both health-related and behavioral factors such as heart rate variability (Sűdy et al., 2019), cardio-vascular disease (Bhar et al., 2022), cognitive performance, age-related changes or different personality traits, etc. (Montaruli et al., 2021; Roenneberg et al., 2004). This incompleteness may be due to the complex, time- and/or money-consuming protocols and the subjective nature of questionnaires applied in the field. In a recent paper, we have already suggested a non-rapid eye movement (NREM) EEG-derived chronotype indicator, which is the nadir of sleep spindle frequency during the sleep period (Horváth & Bódizs, 2024b). Although this serves as a promising indicator of the circadian phase, detecting and analyzing the frequency of sleep spindles are not usually necessary to answer the original research or diagnostic question in routine sleep examination settings. Thus, a metric that can be derived directly from sleep architecture could help bridge the above-mentioned gap. In terms of sleep medicine and research, the main benefit of such a metric is that it eliminates the need for additional devices, multi-day assessments, or resource-intensive protocols beyond the usual PSG recording. Moreover, it offers greater objectivity compared to measurements influenced by individual choices, such as bedtimes or habitual sleep–wake schedules. Furthermore, the utility in determining chronotype from previously collected PSG data could afford a major advantage in retrospective studies.
Nap studies and continuous bedrest experiments suggest that REM sleep is governed by circadian regulation processes. Its distribution throughout a 24-h day is mostly systematic with the highest REM amount in the early morning (Endo et al., 1981). Furthermore, REM propensity has a temporal relationship with the circadian rhythm of the core body temperature (see review in Wurts & Edgar, 2000).
Besides its sensitivity to circadian modulation, the amount of REM sleep was found to be associated with subjective sleep quality in the next morning (Della Monica et al., 2018; Pierson-Bartel & Ujma, 2024). Due to the circadian timing of REM sleep propensity and the relationship between sleep quality and REM sleep duration, REM sleep was proposed as a sleep quality indicator in a literature review (Barbato, 2021). Furthermore, in a recent study, the duration of REM sleep in daytime sleep was found to interact with the adaptation to night shift work, where the group with the larger circadian adaptation had longer REM sleep duration (Zimberg et al., 2024).
As REM amount and REM propensity are the most frequently reported variables related to circadian modulation, we aim to propose a circadian phase indicator that includes both of the above-mentioned aspects of REM sleep. This variable, called the crest of REM propensity (marked hereon as REMmaxprop), is the middle time of the sleep cycle when the ratio of REM sleep is the highest during the sleep period. It encompasses both propensity and duration of REM sleep, as it describes the time when NREM is the shortest relative to the subsequent REM sleep period. Secondly, the accumulation of REM sleep duration prior to the REMmaxprop is also analyzed, as this variable (hereafter referred to as REMacc) is hypothesized to provide insight into both the circadian phase and the association between sleep quality and REM duration described in the literature mentioned above. However, as sleep efficiency/continuity and total sleep time (TST) are the most consistently reported correlates of subjective sleep quality and subsequent sleepiness during the day (see, e.g., in Åkerstedt et al., 2013, 2016; Della Monica et al., 2018; Pierson-Bartel & Ujma, 2024) their interrelationship with REM sleep variables is also an important question.
Overall, the present study aims to examine the relationship between REM sleep and circadian phase, objective sleep quality as reflected in sleep continuity measured by the number and normalized density of arousals, overall subjective and morning sleep quality, as well as morning sleepiness.
The main assumption of the present study is the convergent validity of REMmaxprop as a chronotype indicator. Thus, we hypothesize that it will positively correlate with subjective chronotype, actigraphy-derived sleep midpoint, and with the time of CBT minimum of the participants (meaning later time of REMmaxprop for participants with later subjective chronotype, later time of actigraphy sleep midpoint and later CBT minimum). Furthermore, we hypothesize that both REMmaxprop and REMacc will reflect age-related differences in the circadian rhythm and sleep quality, respectively.
Secondly, we suggest that REM-derived metrics, mainly REMacc and REM percent, will be associated with subjective and objective measures of sleep quality as proposed by the literature discussed above.
Methods
Devices and databases
The Budapest-Munich database consists of 251 (122 females) subjects’ signals recorded in different laboratories (electrodes were placed according to the standard 10–20 system) with participants in the age range of 4–69 years divided into four age groups: children (N = 31, 4–10 years), teenagers (N = 36, 11–20 years), young adults (N = 150, 21–40 years), and middle-aged adults (N = 34, 41–69). A detailed description of the database with information on sampling rates, precision, electrode- and recording locations can be found in Bódizs et al. (2022). In the present study, only EMG, ECG, and EEG C3 channels (for the purpose of automatic arousal detection), along with the manually scored hypnograms of the sleep recordings, were used.
The headband EEG database comprised 90 adults’ data from three different studies (age range 18–59, mean age 25 years). In these studies, all participants slept in their own homes; thus, sleep was recorded in an ecologically valid environment with a Hypnodyne Corp. Zmax EEG headband. The recording device has a sampling rate of 256 Hz at derivations F7-Fpz and F8-Fpz. The three studies included in our current analyses followed different protocols, but each was designed to include at least one night of habitual sleep recorded with a mobile headband, after which participants rated their sleepiness on a 1–10 Likert scale. Additionally, participants completed questionnaires on overall sleep patterns, including the Munich Chronotype Questionnaire (MCTQ) and the Pittsburgh Sleep Quality Index (PSQI), as well as a 5 to 7-day actigraphy recording using the Geneactiv Original wrist-worn accelerometer. Recording devices were the same in all three studies.
The first study involved 45 (24 females) healthy young adults (age range 18–39 years) participating in a 7-day-long examination with a 35-h-long sleep deprivation protocol in the last 2 days. The first 5 days were free of specific instructions and interventions; only actigraphy wearing was required, and baseline sleep of the sleep deprivation protocol was recorded on the fifth night of the week using the EEG headband. This sleep period was self-scheduled and ad libitum with freely chosen bedtimes in the evening, with prohibited alarm clock usage in the morning (more detailed protocol can be found in G. Horváth & Bódizs, n.d.). In the second study, healthy adult participants (N = 23, 14 females) had to follow a regular sleep schedule adapted to their general/habitual routine for 1 week, during which the completion of sleep diaries and sleepiness questionnaires were required before and after the sleep periods. On the sixth evening, participants recorded their sleep with the headband. A subset of these participants (N = 15, nine females) also wore a non-invasive core body temperature (CBT) sensor (CALERA research; sampling rate: 1 Hz) in the first 48 h and a wristband accelerometer (Geneactive Original) during the whole week. Finally, the third study is a larger ongoing research project focusing, among others, on the usefulness of wearable devices in the research of circadian rhythm. According to the study protocol, participants are examined with mobile wearable devices in their everyday life circumstances. In this week-long study, a 7-day wrist actigraphy and a Cortrium C3 + Holter ECG measurement were conducted; furthermore, 3-day-long CBT and EEG measurements were conducted along with morning and evening sleep diaries. From this study, N = 22 participants’ CBT, actigraphy, one night of EEG data, as well as morning sleep quality and sleepiness ratings after the EEG-recorded sleep episode, were included in the present analyses. To sum up the EEG headband database, habitual home sleep records (baseline sleep in the sleep deprivation study), the first 5–7 days of actigraphy measurements (first five nights in the sleep deprivation study ended with awakening from baseline sleep, 6–7 nights for other studies depending on availability), CBT data, sleepiness ratings and sleep quality ratings from sleep diaries from the mornings of the EEG-recorded sleep, MCTQ, and PSQI results were analyzed.
All subjects were free of psychiatric or neurological disorders based on self-reports. In addition, the first two datasets in the wearable EEG database have the same exclusion criteria, including the Hungarian version of the Pittsburgh Sleep Quality Index (Takács et al., 2016) score over 5, Beck Depression Inventory (Beck et al., 1961) score over 12 (moderate and severe depression symptoms; Rózsa et al., 2001), extreme circadian preference (MCTQ chronotype scores outside of the ± 3 SD of reported values in young Hungarian subjects according to Haraszti et al. (2014) and shift work, as well as reported acute and/or chronic medical diagnoses or ongoing pharmacological treatments. The third wearable device study aims to focus on the general population. As a consequence, we only excluded subjects with acute health issues from this subsample of the current report. However, only one participant was (10 min) further than 3 SD from the population mean as indicated by MCTQ results (sleep midpoint range for the three datasets: 01:40:00 AM to 7:35:00 AM, mean 4:19:35 AM) thus, we excluded this participant from the statistical analyses.
The National Public Health Centre Institutional Committee of Science and Research Ethics or the Ethics Committee of the Semmelweis University (Budapest, Hungary), as well as the Medical Faculty of the Ludwig Maximilians University (Munich, Germany), approved the research protocols, and the whole experiment was implemented in accordance with the Declaration of Helsinki. Every participant (or, in the case of minors, their parents or guardians) provided informed consent to participate in the study.
| Sample | Measurements | Questionnaires | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Database | Dataset | Study protocol | Reference | N | Female | Age range (years) | Method | No. of recording days/nights | Analyzed nights (based on availability) | Device | |
| Budapest-Munich | (Bódizs et al.,) [2022] | 251 | 122 | 4–69 | Polysomnography | 2 | 2nd night | Ambulatory PSG | - | ||
| Wearable EEG headband | Study 1 | Sleep deprivation | (Horváth & Bódizs,) [2024a] | 45 | 24 | 18–39 | Electroencephalography | 2 | Baseline sleep | Hypnodyne Zmax EEG headband | PSQI, MCTQ, Likert Sleepiness Scale |
| Actigraphy | 7 | First 5 nights ending with awakenings from baseline sleep (min. 4 nights) | Geneactive Original wristband accelerometer | ||||||||
| Study 2 | Regular sleep schedule with actigraphy and CBT | - | 15 | 9 | 20–42 | Electroencephalography | 1 | All | Hypnodyne Zmax EEG headband | PSQI, MCTQ, Likert Sleepiness Scale (1–10), morning sleep quality (very bad to very good) | |
| Actigraphy | 7 | All (min. 6) | Geneactive Original wristband accelerometer | ||||||||
| Non-invasive core body temperature meas | 48 h | All | Calera research heat-flux sensor | ||||||||
| Regular sleep schedule | - | 8 | 5 | 18–45 | Electroencephalography | 1 | All | Hypnodyne Zmax EEG headband | |||
| Study 3 | General life circumstances | - | 22 | 13 | 19–59 | Electroencephalography | 3 | 1 (the one with best signal quality) | Hypnodyne Zmax EEG headband | PSQI, MCTQ, Likert Sleepiness Scale (1–10), morning sleep quality (very bad to very good) | |
| Actigraphy | 7 | 6 | Geneactive Original wristband accelerometer | ||||||||
| Non-invasive core body temperature meas | Min. 72 h | All | Calera research heat-flux sensor | ||||||||
Analyses of the electroencephalography data
*REMmaxprop is the putative circadian phase indicator of REM sleep proposed in the present study (see example in Fig. 1). Its calculation was as follows:Calculate the duration of REM and NREM periods for every sleep cycle separately.Estimate the ratio of the REM/NREM periods in each sleep cycle by dividing REM duration by NREM duration.Look for the sleep cycle that contains the highest REM/NREM ratio.Determine the middle of the sleep cycle containing the highest REM/NREM ratio and estimate the local time of this cycle-middle from the start time of the sleep recording.
REMand CBT minimum time of a 19-year-old female participant.: CBT data and the fitted sinusoid via cosinor analysis, with the time of minimum indicated by the. Since the curve fitting was applied to the entire recording, the local minima of the cosinor curve are identical across all three nights.: hypnogram of the same subject, where theindicates the estimated time of REM(i.e., sleep cycle midpoint when the ratio of REM sleep is the highest compared to the preceding NREM period) maxprop maxprop Note. Top orange arrows Bottom blue arrow
Subjective sleep quality and circadian phase indicators
Actigraphy-derived weekly average sleep midpoint, time of the CBT minimum, and subjectively measured chronotype were used as validated circadian phase indicators (Reid, 2019; Roenneberg et al., 2019; Santisteban et al., 2018).
Actigraphy sleep midpoints were calculated from sleep onset and wake-up times for all days separately, then the whole week (5–7 days) average of the sleep midpoints was used in the subsequent analyses. Sleep periods, thus sleep onset and wake-up times were estimated with GGIR, an open-source R package (Migueles et al., 2019; Van Hees et al., 2015), using the setting of time window as wake-to-wake.
The CBT signal was subjected to cosinor analysis with the use of an open-source algorithm, CosinorPy (Moškon, 2020). Finally, the time of the first local minimum of the fitted sinusoid on the time series was used as the circadian phase indicator (example in Fig. 1).
Subjective chronotype was derived from the Munich Chronotype Questionnaire (MCTQ), which considers the oversleeping adjusted sleep midpoint (MSFsc: estimated from self-reported bed- and wake times on work- and free days) as a circadian phase indicator (Roenneberg et al., 2003).
Subjective sleep quality was measured in two ways. The PSQI score was used as a measure of general subjective sleep quality. In addition, self-ratings of sleep quality (ranging from very bad to very good on a 1–5 scale) and sleepiness (on a 1–10 Likert scale) were collected in the morning following the night sleep period assessed by the EEG headband.
Statistical analyses
Statistical analyses were conducted with TIBCO Statistica Software. An a priori power analysis was conducted for the determination of the sufficient size of the sample to test our hypotheses. Its results revealed that the required sample size to achieve 80% power for detecting a medium effect (according to Cohen’s guidelines) at a significance level of α = 0.05, was N = 68 for correlational analysis and N = 79 for one-way ANOVA. Thus, the planned sample size of N = 90 (wearable EEG headband database) and N = 251 (BPM database) appeared suitable to test the study hypotheses regarding the associations between EEG and subjective variables.
The two databases were analyzed separately using the following statistical tests. Associations among variables were examined with Pearson correlation (coefficients marked with r in the result section) where the data have Gaussian distribution and Spearman’s rho (marked with R in the result section) was estimated when the data is non-normally distributed, or ordinal (PSQI score, Likert sleepiness, morning sleep quality). Group comparisons were conducted with one-way ANOVA on normally distributed data and with Kruskal–Wallis ANOVA where a non-parametric version of the test was needed.
Results
REM sleep indices, chronotype, and subjective sleep quality (wearable EEG headband dataset)
Since the weekly average total sleep time was significantly longer than TST on the EEG-measured night (mActTSTavr = 7:57 hh:mm, SD = 00:55 hh:mm; mTST = 7:35, SD = 1:09 hh:mm; t(67) = 3.53, p < 0.001) we examined whether this difference in sleep duration between the average and EEG-night was reflected in the subjective quality of sleep assessed in the morning of the EEG-night sleep. We found a tendency for shorter sleep time to be associated with lower sleep quality. Specifically, participants who had shorter EEG-night TST compared to their average sleep duration (greater difference) reported lower sleep quality ratings upon awakening (TSTavr–TSTEEGnight & current sleep quality: r = – 0.33, p = 0.06); however, this was not true for the sleepiness ratings (TSTavr–TSTEEGnightr = 0.15, p = 0.22. We also checked whether REM percent is associated with the difference of TSTavr & TSTEEGnight, as early morning REM loss is highlighted in the literature as a possible cause of sleep quality reduction (Barbato, 2021; Naiman, 2017). We found no significant association between the two variables (REM percent vs. TSTavr–TSTEEGnight: r = – 0.21, p = 0.089).
Associations of REMwith the different circadian phase metrics.Correlation of REMwith MSFsc, weekly average sleep midpoint regarding actigraphy, and with the time of the CBT minimum maxprop maxprop Note.
| EEG | Actigraphy | CBT | MCTQ | Sleepiness scale | Current sleep quality | PSQI | |
|---|---|---|---|---|---|---|---|
| EEG | 80 | ||||||
| Actigraphy | 67 | 76 | |||||
| CBT | 31 | 34 | 34 | ||||
| MCTQ | 73 | 71 | 30 | 83 | |||
| Sleepiness scale | 77 | 74 | 32 | 81 | 87 | ||
| Current sleep quality | 41 | 36 | 34 | 37 | 42 | 44 | 88 |
| PSQI | 78 | 76 | 34 | 82 | 86 | 44 | 89 |
| PSQI | Morning sleep quality | Sleepiness | ||||
|---|---|---|---|---|---|---|
| R | p | R | p | R | p | |
| REMmaxprop | – 0.09 | 0.43 | – 0.1 | 0.52 | 0.01 | 0.96 |
| REMacc | – 0.19 | 0.09 | 0.14 | 0.4 | – 0.2 | 0.09 |
| REM percent | – 0.15 | 0.18 | 0.04 | 0.8 | – 0.1 | 0.37 |
| EEG TST | – 0.18 | 0.11 | – 0.03 | 0.86 | – 0.15 | 0.2 |
REM sleep, age, and objective sleep quality (Budapest-Munich database)
The arousal detection was completed for N = 232 subjects’ recordings, as 19 participants did not have a full-night EMG channel necessary for reliable detection. Other analyses regarding age and REM timing or duration were performed on the whole sample (N = 251). The sample sizes of the age groups were different: children: N = 31, teenagers: N = 36, young adults: N = 131, middle-aged adults: N = 34. A Levene’s test of homogeneity of variances was performed to test the inhomogeneous variances between the different age groups. Age groups were characterized by homogeneous variances regarding REMmaxprop (F(3,244) = 1.55, p = 0.2), but not regarding REMacc (F(3,247) = 5.7, p < 0.01), thus in this latter case the result of Welch’s F-test is reported which corrects for variance heterogeneity.
After normalizing the number of arousals with the length of sleep (arousal count/hour in different sleep phases and during the whole sleep period), correlational analyses revealed no association between TST and any of the arousal density metrics. However, significant negative associations of total and NREM sleep arousal densities with REM percent were revealed. Likewise, higher total arousal density predicted lower REMacc (Table 4). Furthermore, Kruskal–Wallis ANOVA revealed a significant effect of age groups on all three arousal density metrics (H(3)total = 32.4, p < 0.001, η2H = 0.12; H(3)NREM = 29, p < 0.001, η2H = 0.11; H(3)REM = 20.5, p < 0.001, η2H = 0.07).
The timing of REMand accumulation of REM sleep prior to the maximal propensity as a function of age maxprop
| Arousals | Total count | Count in NREM | Count in REM | Total density | Density in NREM | Density in REM | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| r | p | r | p | r | p | r | p | r | p | r | p | |
| REMmaxprop | 0.06 | 0.39 | – 0.04 | 0.58 | 0.07 | 0.31 | – 0.01 | 0.896 | – 0.1 | 0.224 | – 0.06 | 0.408 |
| REMacc | – 0.01 | 0.843 | – 0.03 | 0.62 | 0.26 | < 0.001* | – 0.14 | 0.036* | – 0.11 | 0.086 | – 0.04 | 0.569 |
| REM percent | – 0.12 | 0.065 | – 0.13 | 0.049* | 0.42 | < 0.001* | – 0.21 | 0.001* | – 0.18 | 0.006* | – 0.04 | 0.596 |
| EEG TST | 0.2 | 0.003* | 0.15 | *0.025 | 0.29 | < 0.001* | – 0.1 | 0.159 | – 0.04 | 0.53 | – 0.07 | 0.301 |
Discussion
The present study examines the relationship of REM sleep with the circadian phase as well as with objective and subjective sleep quality. We found that the time of the maxima in REM/NREM ratio (REMmaxprop) reliably reflects the individual differences in the circadian phase and chronotype, the latter measured with both subjective and objective metrics. Statistically significant correlations were found between the proposed circadian phase marker (REMmaxprop) and previously published (questionnaire- and actigraphy-derived) measures, whereas the level of the association between REMmaxprop and CBTmin emerged at a trend level. However, the accumulation of REM prior to REMmaxprop (REMacc) was not associated with measures of the circadian phase, although age seemed to affect this variable (decrease of REMacc with increasing age). Finally, no direct proof of the role of REM-related variables in subjective and objective sleep quality was revealed. Nevertheless, less sleep relative to the weekly average sleep duration tended to be associated with lower subsequent self-reported sleep quality which indicates a within-subject association between the objective and subjective metrics, cohering with earlier reports (Åkerstedt et al., 2013; Pierson-Bartel & Ujma, 2024), as well as a first-night type of effect which could indeed be operative in home sleep studies conducted with self-applicable electrode sets (Miettinen et al., 2018).
Early studies show that REM sleep is promoted by the circadian pacemaker as the crest of its rhythm is preceded by the minimum of the CBT rhythm (Dijk & Czeisler, 1995). However, an easy-to-measure phase indicator was never proposed. REMmaxprop is a promising indicator of circadian timing as it correlates with self-reported chronotype and actigraphy-derived sleep midpoint. Furthermore, the earlier REMmaxprop of children and middle-aged adults as compared to teenagers and young adults must be mentioned. It is well known that chronotype and circadian timing shift to later phases during the transition from childhood to adolescence, whereas this phase delay is followed by a gradual phase-advance with aging (Duffy et al., 2015; Roenneberg et al., 2004). These age effects in the circadian phase align with our results regarding the dependence of REMmaxprop on the age of the subjects. Although REMacc was also affected by age, it seemed to better reflect the age-related changes in REM percent which is reducing during the life span (Floyd et al., 2007; Lokhandwala & Spencer, 2022). Indeed, our results did not support the circadian modulation of this variable, as it was not associated with any of the circadian phase indicators analyzed here.
The present study does not equivocally support the role of REM sleep as an indicator of sleep quality, as no association was found between EEG variables and the different subjective indicators of sleep- and wake quality. However, if we consider arousal count and density as an objective indicator of sleep quality in accordance with several reports in the literature (Zakevicius et al., 2013), REM percent seems to provide information on the integrity of the sleep process, as it correlates negatively with the abundance of NREM arousals. Our findings indirectly suggest that frequent interruptions of NREM sleep by arousals might hinder the emergence of REM sleep. Accordingly, reduced REM duration was found to emerge as one of the most common sleep architectural change across different health problems (Ujma & Bódizs, 2024). Although there were moderate associations between REMacc, REM percent, and the count of arousal in REM sleep, these relationships are confounded by the duration of sleep time as longer TST was also correlated with more arousal events. In spite of the fact that none of the subjective indicators were reflected in the EEG-metrics, a promising picture emerged when the difference of weekly and current sleep duration was compared to subjective sleep quality. The relationship between subjective and objective sleep quality has always been a contradictory question. However, longitudinal studies found within-subject association between subjective sleep quality and the preceding sleep period (Åkerstedt et al., 2013). In a recent study, robust interindividual differences and stable intraindividual dynamics of sleep cycles and sleep stage transitions were reported (Kishi & Van Dongen, 2023). It is likely that the interindividual differences in sleep architecture and sleep duration overshadow the relationship between objective and subjective metrics. Thus, we think that examining “poor sleep” in healthy people requires more than one-night-long sleep recordings performed in ecologically valid settings, which could form the basis for a within-participant analysis of the recordings and self-ratings.
The limitations of the study include the inability to perform circadian analyses and assess sleep continuity within the same sample, and the lack of data on extreme chronotypes, which prevents extrapolation of our findings to those groups. Another limitation is the use of EEG headbands instead of the gold-standard PSG for examining circadian phase. However, this choice also serves as an advantage, given the ecologically valid environment in which the studies were conducted. In sum, our study proposes a new, putative REM-derived EEG indicator of the circadian phase. It encourages further research on the association between subjective and objective sleep quality in longitudinal settings but found no evidence supporting the relevance of circadian regulation – sensitive aspects of REM sleep among the proposed sleep quality indicators. Although the all-night REM percentage partially reflects arousal density, it does not correspond to subjective sleep quality or actual sleep amount relative to the weekly average. Further studies are needed to unravel the relationship between REM sleep and sleep quality.