Effects of a daylight intervention in the morning on circadian rhythms and sleep in geriatric patients: a randomized crossover trial (2024)

A total of 36 patients were enrolled in the study, 18 dropped out. Participant flow is shown in Figure 1(b). Reasons for drop out included withdrawing consent due to the general burden of participating in a clinical trial (n=6), considering the lamp (n=1) or sensors disturbing (n=6), deterioration in health status (n=4) and transfer to another hospital (n=1). There were no adverse events.

Nine subjects were analyzed in both sequence groups, respectively, with one participant per group being discharged prior to completing the second week.

Baseline demographic and clinical characteristics and geriatric assessment are shown in Table 1. There were no significant differences between the two sequence groups.

Two subjects in the C-I group were using mirtazapine as a sleeping aid at study entry and continued using it for the entire period. Three other subjects in the C-I group began taking melatonin as a sleeping aid at 21:00 on day 2, 10 or 13. They were excluded from melatonin analyses. In the I-C group no subjects used sleeping aids during the study.

Table1Baseline demographic and clinical characteristics and geriatric assessment by sequence and by total.

Values are presented as mean ± SD or total number (%). I-C = intervention-control group, C-I = control-intervention group

^aChi-Square test, ^bIndependent two sample t-test, ^cMann-Whitney U test.

Wear time validation

Clinical wear time of the Actigraph was met by 14 subjects (I-C: 6, C-I: 8)., valid wear time by 12 subjects (I-C: 4, C-I: 8).

Variables by period

There were no significant period effects (SI, “tables”, Table SI-1).

Figure 2 shows the mean S-SQ per day depending on the study week for both sequence groups. Linear trend lines show the course of S-SQ per week. In both groups, mean S-SQ decreases during the first week. In the second week, i.e., during the CP, sleep quality increases in the I-C group and even slightly exceeds the initial level of the first week. In the C-I group, sleep quality also increases during the second week, but to a lower degree.

Fig. 2Mean S-SQ in the first week (a) and in the second week (b)

a S-SQ tends to decrease in the first week. The intervention-control group receiving intervention here demonstrated a comparatively smaller decline in S-SQ in contrast to the control-intervention group.

bIn the second week, the S-SQ in the control-intervention group remained almost constant over the course of the intervention. The S-SQ of the intervention-control group, which received no intervention during that week, showed a positive trend.

Cortisol and melatonin by condition

Cortisol levels at 14:00 h were higher than levels at 20:00 h in both conditions. The difference between 14:00 h and 20:00 h cortisol levels was greater in the IP. Neither difference was statistically significant (Table 2).

Melatonin levels were higher at both 14:00 h and 20:00 h and on average during the IP (Figure 3). The difference between midday and evening melatonin levels was also greater during the IP. Neither difference was statistically significant (Table 2).

Table 2Cortisol and melatonin by condition

Values are presented as mean ± SD. ^aPaired t-test, Hedges correction as effect size.

^bWilcoxon's test, Pearson’s r as effect size.

Fig. 3

Comparison of mean melatonin levels at 14:00 h, 20:00 h and on average depending on condition.

Melatonin levels are increased during intervention period.

Whiskers: 95% CI, x: Mean

O-SQ by condition

SE, SOL, and SFI did not increase significantly in the IP. TST, TSTN, and WASO decreased but not statistically significantly in the IP (SI, “tables”, Table SI-2). Although not significantly, the intervention reduced the TSTD from 241 to 202 minutes, corresponding to a reduction of 16%.

S-SQ by condition

The mean S-SQ, restfulness and feeling scores were not statistically significantly higher in the CP than in the IP (SI, “tables”, Table SI-3). Subjective diurnal sleep duration was not statistically significantly shorter during the IP.

Correlation of S-SQ and O-SQ

There was no significant correlation between S-SQ and O-SQ(SI, “tables”, Table SI-4).

DISCUSSION

The aim of the present study wasto determine whether a daylight intervention in the morning could improve the circadian rhythms of cortisol and melatonin and enhance S-SQ and O-SQ in geriatric patients.

Acclimatization to the hospital environment

No significant differences were observed in S-SQ and O-SQ between the first and second week of participants' hospital stay (Table 2, appendix). These findings do not support the hypothesis thatpatients' sleep quality would improve in the course of their hospital stay due to health status improvements[29]. This could be explained by the fact that not all patients exhibited a linear clinical improvement. However, a tendency for S-SQ to deteriorate during the first week and improve again during the second week, irrespective of the condition, could be found (Figure 2). When the CP was in the second week, S-SQ even slightly exceeded the initial level (6.8) at the end of the second period (7.1).These findings suggest a potential period of acclimatization to the unfamiliar environment and subsequent improvement in sleep[4]. The start of the light intervention in the second week requires readjustment to an unfamiliar environment and may have slowed down the overall progress. This may account for the decrease in S-SQ within the C-I group, from 7.0 at the study's beginning to 5.8 at its end.

Cortisol

As hypothesized and in agreement with previous studies, midday cortisol levels were found to be higher in the IP compared to the CP, while evening cortisol levels remained the same[30, 31]. However, these results were not statistically significant.The observed cortisol levels were below the expected values of 2.5 µg/L and 1.9 µg/L for healthy older adults at both 14:00 h, with 1.3 ± 0.8 µg/L (IP, 1.1 ± 0.7 µg/L (CP)) and at 20:00 h with 1.0 ± 0.6 µg/L (IP, 1.1 ± 0.7 µg/L (CP))[23]. Due to the seasonal variation of salivary cortisol, we would have expected high cortisol levels in our study, which was conducted exclusively during winter[32]. It is likely that discrepancies between our findings and comparative values can be explained by the used laboratory techniques. While we measured cortisol using liquid chromatography-tandem mass spectrometry, immunoassays were used in comparative studies. In immunoassays, cross-reactivity can lead to falsely high values[33]. This is not the case with the method used here.

Melatonin

Supporting our hypothesis, mean melatonin levels were higher in the IP than in the CP. The difference in melatonin levels between 14:00 h and 20:00 h increased in the IP compared to the CP. Although the observed differences are not statistically significant, it is noteworthy that the interperiodic difference in the melatonin levels at 20:00 h exceeds 1 ng/L (1.33 ng/L (IP), 0.31 ng/L (CP)). In particular, there is a high variance in the IP. With age, glare sensitivity increases making bright light unpleasant[34]. The actual illumination received by individual participants may have varied due to behavioral reactions related to glare sensitivity, such as repositioning of the lamp and turning away from the light source. Additionally, patients may have inherent issues with light perception, such as those stemming from eye conditions like macular degeneration or cataracts[35, 36]. While all patient files were reviewed for such impairments, it is conceivable that some participants may have had an undiagnosed eye condition.

Similarly to previous studies in geriatric patients[37, 38], this study failed to show significant improvements in circadian rhythms of melatonin, while previous studies have shown beneficial effects of light intervention on melatonin rhythmicity in younger, healthy people[39-41]. This supports that the sensitivity of melatonin to light as a Zeitgeber decreases with age, among other things due to an increasing opacity of the lens[42]. Such a possible insensitivity to light is also reflected in the fact that the melatonin levels at noon are even higher during the IP than during the CP (0.5±0.6 ng/L vs. 0.2±0.3 ng/L), whereas insufficient daylight is thought to increase diurnal melatonin levels[43].It was observed that melatonin levels remain below expectations at all times. It has been shown previously that there is a seasonal variation in melatonin concentration with particularly low levels in winter[44]. However, the levels observed with mean melatonin levels at 20:00 h of 1.3 ng/L (IP) and 0.3 ng/L (CP) are significantly lower than the levels of 2.1 ng/L in the mean of November and February at 20:00 h in older people expected from previous studies[44]. This could be because our subjects were complex geriatric patients rather than older participants with no health difficulties. In non-healthy older people, a decrease in the mean melatonin concentration has been shown with an increase in diurnal melatonin levels[43, 45, 46]. Even with this approach, our results are significantly lower than in previous studies in similar subjects. As with the low cortisol levels, this could be explained by the fact that immunoassays were used in comparative studies, where cross-reactivity can lead to falsely high levels, whereas we used liquid chromatography-tandem mass spectrometry.

It is noteworthy that three participants in the C-I group were prescribed new sleep medication during the study period. Melatonin was given to all three patients at 21:00 h, it was administered to two of the three patients only in the middle or towards the end of the second week. None from the I-C group were administered new sleep medication. This variation among the sequence groups could suggest that the light intervention was effective in terms of stabilizing the rhythm, as none of the participants who had the IP first required additional sleep medication.

S-SQ

Previous studies on the effects of a light intervention on S-SQ are inconclusive, with none showing a significant improvement[9]. In contrast to our hypothesis that the daylight intervention would improve S-SQ, a non-significant deterioration from 6.3 at day 1 to 6.1 at day 6 was observed during the IP. As described, behavioral reactions (turning away from light source due to glare perception) and physiological conditions (opacity of lens and thus reduced actual signal) in combination can lead to likely smaller effect of the intervention. Previous studies have shown that the subjective assessment of sleep is heavily influenced by surrounding conditions and how individuals cope with their current situation[14]. Good sleep quality can be observed when individuals adjust to their illness and hospitalization, while bad sleep quality is associated with a sense of helplessness[14]. The hypothesis that not directly sleep-related factors strongly determine S-SQ could explain the lack of a positive trend of the S-SQ in contrast to objective parameters. Nevertheless, S-SQ is also an important parameter to record, as it allows conclusions to be drawn about the overall condition of the patient and is even an independent predictor of mortality in older adults[47].

O-SQ

The actigraphy results were inconclusive and lacked statistical significance. Nevertheless, the intervention resulted in a clinically significant decrease of TSTD from 241 to 202 minutes.In contrast to previous studies and our hypothesis, no increase in TST and TSTN could be observed during the IP[37, 38, 48]. Unexpectedly, a high SE could be observed (97.8 ± 1.9 % (IP), 97.0 ± 2.0 % (CP))[14, 49].While wrist-worn actigraphy estimates objective sleep parameters almost correctly compared to polysomnography, hip-worn actigraphy overestimates SE by more than 13% and TST by approximately 14%[50]. If we assume that our results are biased by the same factor due to the measurement method, we can report a value of 86.1% and 85.3% for SE in the IP and CP, respectively and a TST of 339.7 minutes (IP, 350.9 minutes (CP)). Considering the measurement method, our results are consistent with previous studies in similar populations[14, 49].

Correlation of S-SQ and O-SQ

In line with our results, previous studies have shown that the discrepancy between S-SQ and O-SQ increases with age, and that there is little correlation, especially if they already report sleeping problems[14, 15]. As previously stated, the discrepancy may be attributed to the notion that S-SQ depends less upon the actual sleep and more upon the manner in which individuals handle the situation through coping mechanisms and psychological factors.

Limitations and Strengths

It should be noted that this study also has some limitations.

First, the sample size of 18 participants limits generalizability. Although the sample collected was considered sufficient to demonstrate potential light intervention effects, limitations such as missing saliva samples due to COVID-19 and with data collection methods, did not allow the power needed for the analysis.

Second, exposure to the daylight lamp for 5 hours for 6 days during the IP could not be ensured. With increasing sensitivity to glare with age,participants might have felt negatively affected by the lamp possibly leading to behavioral countermeasures. These behaviors were mitigated through regular rounds and adjustments to the lamp with the subjects. In future research, it may be preferable to use non-adjustable light sources and allow subjects to adjust the illuminance themselves, thereby increasing actual exposure by reducing behavioral countermeasures.

Additionally, data from the light sensors could not be used for the main analysis due to methodological limitations that reduced the quality of the data. Further explanation is provided in the supplementary material “Illuminance”.

Conducting studies with acutely ill geriatric patients poses specific challenges that require addressing. Deteriorating health conditions may increase the likelihood of high dropout rates. Furthermore, crossover studies are known to be particularly challenging for patients, as the observation period per subject is twice as long as in a parallel group design[28].Consequently, study personnel have been intensively deployed to mitigate this challenge. This could not completely overcome the additional burden of study participation in the already stressful situation of being ill and hospitalized.

On the other hand, the crossover design is also a particular strength of the study. The patients served as their own controls. The dependence of the treatment effects on underlying diseases and sleep problems can be neglected in this study design, in contrast to a parallel group design. In addition, changes in circadian rhythms, showing large inter-individual differences in older and hospitalized people, can be better compared. It is also possible to assess S-SQ and to compare it between study periods. Adaptive randomization minimized possible effects of the order of the interventions. The one-day washout period minimized the probability of carry-over effects.