Various physiological processes and medical conditions in humans are known to exhibit seasonal patterns. This also includes medical conditions that affect sleep. Insomnia is a highly prevalent disturbance in the general population (Schlack et al., 2013) but results on its seasonal pattern are unclear. On the one hand, a study comparing weekly self-reported measures between January and August in Norway (69°N latitude) found seasonal differences in insomnia (Friborg et al., 2012), whereas, on the other hand, a large (2,614 participants) epidemiological study in Japan (36°N) did not find an association between insomnia and seasonality (Itani et al., 2016). Another common sleep-related disorder affected by seasonality is obstructive sleep apnea, reported to increase in winter and decrease in summer (Cassol et al., 2012). Naturally, seasonality is evident in seasonal affective disorder (SAD) including lengthened sleep and weight gain during winter (Rosenthal et al., 1984), which is more prevalent in people living at higher latitudes (Rosen et al., 1990). SAD patients experience rapid positive effects of light therapy (Rosenthal et al., 1984), with effects being more pronounced using light of high intensity, long exposure early in the day, and exposure to the eyes rather than the skin (Wehr and Rosenthal, 1989).

Light exposure is known to impact sleep architecture (Dijk et al., 2012; Chellappa et al., 2013; Wams et al., 2017). In healthy humans, the influence of the length of scotoperiods (periods of darkness) and melatonin secretion was already demonstrated in laboratory settings. Participants’ sleep period time (SPT) became 3.3 h longer under 10 h photoperiods (period of light) than under 16 h photoperiods (Wehr, 1991). The most important factor for changing sleep durations, apart from sleep debt, might therefore be naturally changing light exposure throughout the year. The length of photoperiods and scotoperiods inversely vary over the year due to the skewed rotation axis of the earth relative to its plane of orbit around the sun. Light plays a significant role in adjusting to seasonal changes since light input to the retina is the main zeitgeber (time cue) for the biological clock. Circadian entrainment of the biological clock is mediated by the suprachiasmatic nucleus (SCN), which controls the secretion of melatonin by the pineal gland (Czeisler, 1995; Maywood et al., 2021). Melatonin is then suspected to synchronize the clocks in the cells of the periphery (Lewy et al., 1992). The hormone is secreted exclusively during scotoperiods (Wehr, 1991) and was shown to vary across the seasons in peak concentration (Adamsson et al., 2017) and duration of secretion (Kauppila et al., 1987). In addition, the SCN itself has been shown to vary with seasons in its volume and in the number of vasopressin-immunoreactive neurons it contains (Hofman and Swaab, 1992).

While biological processes and findings in laboratory settings would speak for seasonal changes in human sleep duration throughout the year, the expression of such changes in real life and especially in urban settings is contested. In the 1990s no difference in sleep timing and length was found when controlling for the midpoint of the nocturnal period (Wehr et al., 1995). A substantial amount of work had shown inconsistent results on seasonal variation of melatonin levels over the year for humans living in an urban environment (Wehr, 1997). Thus, the function of melatonin on human seasonality was questioned in general. The recently reported association of subjective seasonality and subjective sleep duration with individual melatonin levels determined by the degree of pineal calcification (Kunz et al., 2021) may have provided the missing link. In that study, the subjective sleep duration of healthy adults varied by approximately 1 h over the year. The next step is to investigate seasonal variations in objectively measured sleep parameters.

Investigating the sleep architecture of large patient groups, while not transferable to assessments of seasonality in a healthy population, would allow us to obtain some preliminary observations. Previously, variations were found between seasons in rapid eye movement (REM) sleep duration and REM latency for men with erectile problems and some with suspected sleep apnea in the subtropical climate of Israel (Askenasy and Goldstein, 1995). REM duration was shorter in spring and summer and longer in autumn and winter. REM latency, on the other hand, was longer in autumn and winter. However, these findings were not replicated in a group of 706 unselected apnea patients (Herer and Lavie, 1997). Both studies examined transient variations by looking at changes in sleep parameters grouped by season. Of special interest are REM sleep parameters, since REM sleep is known to be mainly under circadian regulation (Endo et al., 1981; Dijk and Czeisler, 1995; Bes et al., 1996; Khalsa et al., 2002), being entrained by the daily light-dark cycle.

Our study aimed to determine objective sleep duration and architecture over 1 year in a large group of patients attending a neurologic/psychiatric sleep laboratory. In comparison to other similar studies, patients’ usual environment was not manipulated before the polysomnography (PSG) recording and PSG data were analyzed per month.

Monthly means of sleep parameters with SD, number of patients per month, and LMM results are presented in Table 2. Figures 1, 2 show the MvA of TST, REM sleep, and SWS. The Ljung-Box Tests indicate the existence of significant serial correlations (p < 0.001) for all sleep parameters.

The LMM was not significant for TST, SPT, sleep latency, or sleep efficiency. The monthly mean of the TST reaches a maximum difference between January and June of ∼62 min. Its MvA displays a clear pattern influenced by the seasons, with longer sleep in winter and shorter in spring. Between January and May, the patients’ MvA of TST decreased by 43 min. Though not displayed in Figure 1, SPT shows a similar pattern, losing 55 min of SPT between January and June. For both parameters, as well as sleep latency, the MinLag occurred at 27 weeks. This is very close to half a year (26 weeks). Sleep latency had its peak in January, but did not reach its trough till October (ΔMvA ≈ 9 min). Sleep efficiency fluctuated much stronger than the other parameters (MinLag = 14 weeks) but only by up to 3.6%.

Three of the four REM sleep parameters displayed a significant main effect of “month of record.” In September, December, and January the amount of REM (min) (F_11,176 = 2.39, p = 0.009) is relatively high (monthly means ≈ 100 min) and in April and June low (monthly means ≈ 70 min). A similar time course is displayed by REM (%) (F_11,176 = 2.31, p = 0.011), with its highest monthly means in December and January at ≈ 22% and lowest monthly means in April and June at ≈ 19%, see Figure 2). REM (%) and REM (min) both have a near identical MvA pattern and a MinLag of 27 weeks.

Rapid eye movement latency also showed a significant main effect of “month of record” (F_11,176 = 2.34, p = 0.010). It peaks twice, in May and November, and has troughs in January and September. The fluctuation between the seasons can be seen in its MvA pattern with a MinLag of 14 weeks (close to the 13-week length of a quarter of a year).

No significant difference in variation was found in REM stability between months. Its MvA pattern peaks in February, dips to its lowest in September (ΔMvA = 2.25 min), and came with a late occurring MinLag (29 weeks).

Slow wave sleep (%) showed a significant main effect of “month of record” (F_11,176 = 2.19, p = 0.017). Mostly stable throughout the year with peak values from December to May, SWS was significantly reduced in September. While SWS (min) was not significant, its MvA displays a difference of 31 min between September and February. The MinLag of SWS (min) occurred late at 29 weeks. This reflects the steep fall and rise pattern of SWS during autumn. Furthermore, SWS (%) seemingly displays an inverse relationship to REM (%) exhibiting a clear trough in September and October (see Figure 2).

Linear mixed-effects model showed no significant effect for N2 (min and %) but displayed a similar MvA pattern to TST with peaks in November and troughs in May (MvA = 231 min and 197 min; MinLag = 27 weeks). The LMMs of arousal index, awakening index, and PLM index showed no significant effects. The MvA patterns of arousals and PLM index fluctuated strongly (MinLag = 14 weeks), while the fluctuation was less pronounced for awakening index (MinLag = 16 weeks).

To examine a possible confounding effect of insomnia diagnosis on sleep duration, two independent t-tests were performed on TST between the astronomical seasons of spring and summer with winter. There was a significant difference between spring and winter TST (U_39,42 = 638; p = 0.023; d = 0.24) with a medium effect size and 25.5 min longer TST in winter. Both seasons contained the same number of patients with insomnia (N = 21).

Data from our study obtained from a heterogeneous patient group with sleep-related disorders show a significant variation of REM-minutes and -percentage of TST, with more REM-sleep during winter and less REM-sleep during summer. REM sleep latency was significantly shorter in winter and early autumn compared to late spring and late autumn. SWS shows a rather stable course over the year except for a significant decrease in autumn.

Looking at REM sleep parameters, the present finding confirms a previous observation of significant changes occurring in REM duration between seasons (Kohsaka et al., 1992; Askenasy and Goldstein, 1995). It has been demonstrated that higher-intensity light exposure suppresses subsequent REM (%) (Wams et al., 2017). Longer photoperiods, which are common in summer, are related to shorter duration of melatonin secretion (Wehr, 1991) and under natural light conditions (camping settings) were shown to be linked to delayed melatonin onset (Wright et al., 2013; Stothard et al., 2017). This results in stronger entrainment and later dim light melatonin onset, which has previously been shown to accompany decreases in REM sleep (Cajochen et al., 1997; Gordijn et al., 1999). The significant reductions in REM sleep in April and June could therefore be related to stronger circadian entrainment in summer (Roenneberg et al., 2007; Zerbini et al., 2021). In winter, entrainment effects are said to be reversed due to insufficient light exposure (Adamsson et al., 2017), which fits the observed significant increase in REM duration in January.

Since SWS is generally under homeostatic control (Borbély, 1982; Dijk et al., 1989), a longer daily wake time in summer should at most lead to deeper, and possibly longer SWS. Previously a 5.5% decrease in SWS (% of SPT) in winter compared to summer has been observed (Kohsaka et al., 1992) and SWS (% of TST) was found to accumulate higher in participants exposed to higher maximum light intensities (Wams et al., 2017). A significant decrease of SWS (% of TST) during autumn was therefore unexpected. The underlying factors that lead to these findings can only be speculated on.

This is one of few studies investigating seasonal changes in PSG-documented human sleep architecture that looks for subtle changes by comparing months and not just seasons. While the present study did not find a significant difference in objective sleep duration from the PSG-recorded data measured by month, a decrease in TST is evident from the MvA displayed in Figure 1. Monthly comparison enables one to identify more precise temporal changes. For example, in another study, there seems to be a clear increase in REM (min) between October (M ± SD ∼ 46 ± 4 min) and November (∼ 54 ± 4 min) that might be masked in their statistical analysis, since both months are pooled in their regional definition of autumn (Askenasy and Goldstein, 1995). However, the trade-off for higher precision, when using monthly comparisons instead of comparing seasons, is a reduction in statistical power. Parameters of sleep duration and sleep architecture display very large SDs, indicating a high variability between patients. On average the monthly TST SD is 71.5 min, compared to the maximum TST monthly mean difference of 61.7 min. Therefore, the lack of significant variance between months might be due to an insufficient amount of participants.

Recording patient data in a neurologic/psychiatric sleep laboratory has the advantage of allowing patients to both choose their “light off” time and to be able to “sleep in,” which can be argued to be essential for measuring sleep duration. The patients in this study slept substantially longer than in comparable studies (Askenasy and Goldstein, 1995; Herer and Lavie, 1997). One of those authors reported that the patients were awakened (Herer and Lavie, 1997), and it can be assumed that this happened in both studies. This is the norm in most sleep laboratories, especially those focused on internal medicine, where patients are generally scheduled at a set time and sent to bed once the setup is complete. On the other hand, the present study was limited by the variety of diagnosed patients with different diseases and the various intakes of different medications.

The benefits of investigating available data from a large and diverse group of patients must be weighed against collecting data from a most likely smaller group of selected healthy participants. While the latter would be a preferable scenario, the former enabled one to quickly analyze a substantially large participant pool that is mostly spaced out evenly throughout the year. The sleep architecture of insomnia is marked by prolonged sleep latency and strongly reduced TST, but the perception of reduced sleep tends to be some more severe than PSG data indicates, likely/possibly due to an increase of “micro-arousals” and alpha wave intrusion (Erman, 2001). Studies investigating the seasonality of insomnia either find no changes between seasons (Itani et al., 2016), or an increase of insomnia in winter (Husby and Lingjaerde, 1990; Friborg et al., 2012). The fact that the dataset shows a seasonal difference, as opposed to the expected changes, could indicate that these changes may be even greater if generalized to a healthy population. Nevertheless, it was important to carefully account for confounding factors in the patient group. Since certain medications, that are prevalent to be taken by a neuropsychiatric cohort, interfere with sleep in multiple ways, a large number of patients had to be excluded from the analysis. Potentially, further research could be conducted by analyzing multiple years of sleep data to replicate these findings and to increase monthly group size.

Seasonal changes in sleep architecture may have implications for recommendations regarding sleep routines. For many people, the time to wake up is more strongly controlled by their employer’s business hours or school times than by their internal clock. Adjustment of sleep schedule can, therefore, only be controlled by choosing the time to go to bed. Nevertheless, keeping to the same time is widely recommended and often enforced religiously on children. Our findings suggest that improvements can be made by accounting for the increased sleep need in winter, by going to bed earlier. Furthermore, the current twice-annual turning of the clock may facilitate adjustments to seasonal changes in human physiology underlying the findings we report here. For example, standard time in winter allows for more natural light exposure during morning commutes. On the other hand, the reduced sleep need in summer can result in longer periods of wakefulness prior to school or business hours that could be better spent during spare time in the afternoons and evenings. Especially younger people and women potentially benefit from daylight saving time this way since they were shown to exhibit stronger seasonality (Kasper et al., 1989).

In conclusion, the observed significant change in prolonged REM sleep duration in winter vs. spring was in line with previous studies. The data indicate that sleep architecture in patients with sleep disorders is affected by seasonal changes, even in an urban environment with low natural light exposure and high light pollution. We do not intend to state hard facts about the severity of changes in sleep duration between seasons in the general population but wanted to highlight that changes can be observed in a sizeable subpopulation. Since the 90s, the research on the seasonality of sleep architecture has halted. While experimental differences in sleep architecture in response to changing light exposures are well documented, the effect of season on human sleep is thought to be negated by our unlimited access to light and temperature sources, especially in an urban environment. Previously we discovered significant changes between seasons in the subjectively reported sleep duration of a homogenous group (Kunz et al., 2021). In our retrospective study, we were able to obtain an unusually large dataset for a study in this field. Being able to demonstrate these differences objectively in a heterogeneous patient group is of value for sleep schedule recommendations to patients and can hopefully be used as an incentive to justify larger studies on this topic again. Furthermore, it bears investigation if these changes can be directly related to underlying factors, such as sunshine duration and temperature changes and if the observation of a decrease in SWS in autumn can be replicated for further years. Overall, potential benefits of adjusting sleep behavior with the changing seasons seem evident, at least for individuals with insomnia.

The original contributions presented in this study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. The patients/participants provided their written informed consent to participate in this study.

JZ, FB, and DK designed the experiment. AS performed the study and drafted the article. JZ, AS, and KW analyzed the data. AS and KW prepared the figures. All authors edited and reviewed the article, and approved the final version.