Promoting Healthy Sleep–Wake and Circadian Rhythms to Prevent Delirium: Next Step, Target Engagement

Sleep–wake and circadian rhythm disturbances have long been hypothesized as a cause of delirium. Notwithstanding research from the 1950s on the effects of prolonged sleep deprivation [1], data on the extent to which sleep disturbances in hospital settings contribute to delirium remain inconclusive. To test this putative mechanism, modern studies have investigated whether pharmacological and non-pharmacological interventions that promote healthy sleep–wake and circadian rhythms might prevent delirium. In this issue, de Oliveira et al. present a systematic review of the literature to answer the question: Do melatonin agonists or dual orexin receptor antagonists (DORAs) reduce delirium incidence? [2]. The authors include evidence from both randomized controlled trials (RCTs) and observational studies, which are analyzed separately. They then use meta-regression to identify potential differences in clinically relevant subgroups. First, based on pooled results of RCTs, the melatonin agonists melatonin and ramelteon were not found to reduce delirium incidence (RR 0.94, 95% CI 0.72–1.22, k = 16 studies; RR 0.63, 95% CI 0.39–1.03, k = 8, respectively). However, pooled results from 3 RCTs found that the DORA suvorexant significantly reduced the incidence of delirium versus placebo (RR 0.55, 95% CI 0.35–0.87). In observational studies, which are susceptible to several sources of bias such as confounding by indication and selection bias, the authors again found that melatonin was not associated with lower delirium incidence (k = 3). However, in observational studies, both ramelteon and DORAs were associated with a statistically lower incidence of delirium relative to comparison groups (four studies each). The only significant finding in subgroup analyses was that, in observational studies, a greater effect was found in study arm subjects in surgical-only cohorts than in medical or mixed cohorts. When interpreting the results of this review, it is important to differentiate sleep–wake and circadian rhythms, as these two are related but distinct and dissociable biological constructs. Sleep–wake refers to canonical states of sleep and wake whereas circadian rhythms are endogenous physiological cycles of roughly 24-h duration, which can be assessed reliably only by way of physiological measures like the timing of dim light melatonin onset (i.e., when melatonin levels rise in the evening due to fading light of day, often evaluated in saliva) or the nadir of core body temperature [3]. Moreover, circadian rhythms are not observable at the bedside because they are commonly masked by behavioral and environmental factors [4]. Regarding effects on sleep–wake, DORAs promote sleep by blocking the activity of the wake-promoting compound orexin. Melatonin agonists, however, have modest if any effect on sleep [5]; they are classified as chronobiotics as their primary effect is to shift circadian phase as described by the phase response curve (Figure 1). This review's null results for melatonin, both in RCTs and observational studies, question the role of melatonin in preventing delirium. The discrepant results between RCTs and observational studies for ramelteon are less clear. It could be that there is a true but weak signal in favor of ramelteon, or the results from observational studies could be a false positive due to confounding by the limitations of observational study designs. In either event, the results do not support the routine use of either melatonin or ramelteon for delirium prevention. Highlighting different perspectives in the field, recent clinical guidelines point in different directions regarding the use of melatonin agonists for delirium prevention. In the Focused Update to the PADIS Clinical Practice Guidelines [8], the Society of Critical Care Medicine issued a conditional recommendation suggesting the use of melatonin in adult patients in the ICU. This was based, in part, on low certainty evidence that melatonin may reduce delirium incidence. Conversely, the American Psychiatric Association's Practice Guideline for the Prevention and Treatment of Delirium [9] suggests that melatonin and ramelteon not be used to prevent or treat delirium. This was based on inconsistent and low-quality evidence. The current results [2] are more in line with the latter of these two positions. Several methodological limitations across included melatonin agonist studies should be acknowledged (see Table 1, adapted and modified from the original authors' eTable 1). The primary limitation is most studies' lack of critical engagement with circadian science. No study evaluated subjects' circadian phases to inform the optimal timing of melatonin. When doses are mistimed relative to a person's circadian phase, they might yield no or even counterproductive effects, contributing to misalignment between circadian phase and sleep period. In fact, many studies did not specify timing of administration, which is striking because the timing of administration is essential to the phase-shifting effect of melatonin agonists. Moreover, three studies provided melatonin agonists only before surgery yet evaluated delirium after surgery, and one administered melatonin in relation to the time of surgery independent of the time of day. Second is the large variability of doses across studies. There is a 20-fold difference between the lowest (0.5 mg) and highest (10 mg) dose of melatonin administered. Even a 6-fold difference in melatonin dose has been shown to have different phase response curves [6]. Third, objective circadian outcomes were not evaluated, so we do not know which if any of these studies engaged their purported target. For instance, in a separate analysis from one included study, Jaiswal et al. found no differences in circadian measures of actigraphic activity between ramelteon and placebo arm participants [10]. Although this suggests that ramelteon did not exert a circadian effect in this study, rest-activity patterns are only a proxy for circadian phase. Finally, only two of the seven studies that evaluated subjective sleep reported statistically better sleep among melatonin agonist arm participants. Evidence both from RCTs (k = 3) and from observational studies (k = 4) find a preventive effect of DORAs on delirium incidence. All three RCTs studied the same agent suvorexant as did three of the four observational studies. One observational study included both suvorexant and lemborexant, and no included studies investigated daridorexant, although a small trial with this agent was recently published [11]. The pooled results for DORAs are promising, but the authors are appropriate to regard these findings as hypothesis-generating given the small size of most included studies. As a minor point, the authors use the term orexin receptor antagonists to refer to DORAs, but selective orexin 2 antagonists have also been developed (e.g., seltorexant) though not yet studied in relation to delirium. There are notable methodological limitations across DORA studies as well. First, certain aspects of each of the three RCTs deserve consideration. For one, readers may be skeptical of the reported 0% delirium incidence in the suvorexant arm in the earliest published RCT [12]. In the RCT published the following year, 80% of the comparison arm participants received as-needed trazodone, introducing a potential confound [13]. The most recent RCT, from 2024, enrolled 203 participants across 50 sites, raising questions about interrater reliability; it also restricted eligibility to individuals with mild cognitive impairment, mild dementia, or history of delirium, thereby constraining generalizability [14]. Second, the DORA studies have not investigated objective sleep outcomes as a demonstration of target engagement, though the results of the recent REPOSE study, which does evaluate objective sleep, are forthcoming [15]. Third, all included studies enrolled exclusively Japanese participants, so their generalizability to other populations is uncertain. Nevertheless, the consistency of results across studies is promising and calls for high-quality RCTs of this class of agents to establish efficacy. Sleep interventions offer a variety of novel opportunities in relation to delirium prevention and management [16]. For instance, even stronger than the effect of melatonin agonists on circadian rhythms is the effect of timed exposure to bright light (Figure 1). In fact, early evidence suggests that timed lighting might have a role in delirium prevention [17]. Regarding sleep aids for delirium, especially in older adults, the goal is to identify agents that are both effective and well tolerated, agents without GABA-ergic and anticholinergic activity. Beyond DORAs, alpha-2 adrenoreceptor agonists such as dexmedetomidine have been shown to improve sleep at night with limited effects on sleep architecture [18], and their use exclusively at night may also reduce delirium [19]. Conversely, interventions that promote wakefulness and improve attention and alertness during the day, such as caffeine, also represent novel lines of investigation for delirium prevention [20]. Whereas the findings of this review do not support the routine use of melatonin agonists or DORAs for delirium prevention—nor do they address the question of whether these agents might be efficacious in treating delirium—they may be seen as a call for methodologically rigorous RCTs for these and other sleep–wake and circadian interventions. Future studies of melatonin agonists should evaluate circadian phase, both to guide optimal timing of administration and to demonstrate circadian engagement. Future studies of DORAs that objectively evaluate sleep would also serve to establish the mechanistic link between improved sleep and delirium prevention. Ultimately, it would be refreshing, like a good night's sleep, to find that sleep–wake and circadian interventions not only clear the mind but preserve cognition as well. The author is solely responsible for this article. The work is supported by the National Institute on Aging of the National Institutes of Health under Award Number K23AG072383. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This work was supported by the National Institute on Aging, K23AG072383. The author has nothing to report. The author declares no conflicts of interest. This publication is linked to a related research article by Oliveira et al. To view this article, visit https://doi.org/10.1111/jgs.70305.