Dementia is a common presentation in clinical practice, and its prevalence is expected to increase (Besley Reference Besley2023). Sleep disturbances, including increased wakefulness, sleep latency, daytime sleepiness and fragmented sleep, are common in dementia, and a bidirectional relationship has been suggested between sleep disturbance and amyloid deposition (Holth Reference Holth, Patel and Holtzman2017; McCleery Reference McCleery and Sharpley2020).
Sleep disturbance is of clinical importance as it is associated with agitation, aggression, increased falls, and carer stress and burnout (Stone Reference Stone, Blackwell and Ancoli-Israel2014; Cipriani Reference Cipriani, Lucetti and Danti2015; Brewster Reference Brewster, Wang and McPhillips2024). Those with sleep disturbance are more likely to require earlier institutional care (Lee Reference Lee and Thomas2011).
There is limited evidence to support the use of pharmacological interventions for sleep disturbances in dementia (McCleery Reference McCleery and Sharpley2020). Their use, particularly the use of antipsychotics, is associated with numerous risks (Díaz-Gutiérrez Reference Díaz-Gutiérrez, Martínez-Cengotitabengoa and Sáez de Adana2017; Rochon Reference Rochon, Vozoris and Gill2017; Mok Reference Mok, Carr and Guthrie2024). Given this, there is clearly a potential role for non-pharmacological approaches to sleep problems in dementia. This is reflected in current National Institute for Health and Care Excellence (NICE) guidance (NICE 2018).
A range of non-pharmacological sleep interventions, such as light therapy, behavioural measures and structured programmes, exist but there is no current consensus as to which is most effective (Blackman Reference Blackman, Swirski and Clynes2021).
The Cochrane Review
Outline of the study design
The review in this month’s Cochrane Corner, by Wilfling et al (Reference Wilfling, Calo and Dichter2023), aimed to evaluate the benefits and harm of non-pharmacological interventions for sleep disturbance in dementia and to describe the components of any complex interventions.
Randomised controlled trials (RCTs) that compared any non-pharmacological intervention with usual care, another non-pharmacological or pharmacological intervention and reported a sleep-related primary outcome were eligible for inclusion. There were no language restrictions.
Eligible studies were identified from the Cochrane Dementia and Cognitive Improvement Group’s (CDCIG) Specialised Register (which includes studies identified from healthcare databases), trial registers, the Cochrane Library’s Central Register of Controlled Trials (CENTRAL) and grey literature sources from the Web of Science Core Collection (Box 1). Study authors and field experts were contacted for ongoing or unpublished trials, and references of relevant studies were reviewed.
Grey literature
Grey literature includes material that is not controlled by commercial publishers, and might include theses and dissertations, government reports, policy documents, conference papers, ongoing or unpublished research and personal correspondence (Paez Reference Paez2017).
Grey literature-specific databases do exist, such as the Web of Science Core Collection (Paez Reference Paez2017), which was used in this review (Wilfling Reference Wilfling, Calo and Dichter2020). There are also alternative web search engines, such as Mednar, or material can be manually sourced by searching conference programmes or contacting relevant individuals for unpublished work (Paez Reference Paez2017). Using grey literature helps to avoid publication bias (this refers to the phenomenon by which studies with positive results are more likely to be submitted for publication, accepted for publication, published in high-impact journals and in English) (Conn Reference Conn, Valentine and Cooper2003; Hopewell Reference Hopewell, McDonald and Clarke2007). Not using grey literature can risk overinflating positive results, and grey literature may provide an otherwise unseen side of the evidence (Conn Reference Conn, Valentine and Cooper2003; Paez Reference Paez2017).
Identification of eligible studies, data extraction and assessment of bias were reviewed independently by two review authors.
People with a diagnosis of dementia based on any established diagnostic criteria, or a Mini-Mental State Examination (MMSE) score of less than 24, were included. There were no restrictions on age, setting, type or severity of dementia. Those with sleep apnoea were excluded. Studies with mixed samples were included if at least 80% of participants had dementia. Participants were required to have a sleep problem at baseline, defined by any subjective or objective measure.
Primary outcomes were objective measures of sleep, measured using technology or direct observation, and adverse events. Secondary outcomes included subjective sleep-related outcomes, adherence to the intervention, intervention acceptability, quality of life, functional status, and behavioural and psychological symptoms of dementia (BPSD).
Continuous outcomes were expressed as mean differences, and dichotomous outcomes as risk ratios (RRs), both with 95% confidence intervals (CIs). For interventions not expected to have a prolonged intervention effect, the data point directly after the intervention period was used. For interventions anticipated to need longer follow-up to be fully implemented, the last available follow-up data point was used.
Interventions were grouped into categories. Only the social activities category was sufficiently homogeneous to allow meta-analysis, using a random-effects model.
Summary of the identified studies
The initial search identified 11 278 records. Following removal of duplicates, first assessment by the CDCIG’s Trials Search team, and review authors’ screening of titles and abstracts, 119 full-text articles were reviewed. After exclusions, 19 studies were included in the qualitative synthesis, 2 of which were also subjected to meta-analysis.
Of the 19 studies, 13 were conducted in nursing homes, 5 in the community and 1 in an in-patient setting. Most of the studies (n = 14) were conducted in the USA, and the remainder were in China, Switzerland and Japan. For the comparator conditions, 8 studies used usual care as a control; 6 of the light therapy studies used controlled lighting; 1 study used placebo transcranial electrostimulation; and other comparators included noise reduction and non-directive staff support.
A deeper dive into the available studies
Light therapy
Seven studies (318 participants) examined the effect of light therapy as a stand-alone intervention. With light therapy, nocturnal sleep time ranged from 33 min shorter to 110 min longer; sleep efficiency ranged from 4.5% lower to 16.6% greater; mean night-time total wake time was 39 min lower to 32 min higher; night-time awakenings were 2.9 fewer to 4.89 more; and sleep onset latency was 6.05 min higher to 3.72 min higher.
Physical activity
Three studies examined the impact of physical activities, although there was no clear explanation as to why results were only presented for two of these RCTs. Of the reported studies (167 participants), both reported nocturnal sleep time as 11.8 min longer and sleep efficiency 2.6–4.9% greater with physical activity. One study reported total night wake time as 33.2 min lower, with 3.3 fewer nocturnal awakenings. One study reported no difference in sleep onset latency, but gave no supporting data. One study reported adverse events, including chest pain, electrocardiogram (ECG) changes and musculoskeletal pain.
Social activities
The two studies (236 participants) that investigated social activities were sufficiently similar to allow meta-analysis. The mean differences in total sleep time and sleep efficiency were 16.78 min higher (95% CI −7.78 to 41.34) and 2.65% higher (95% CI −1.79 to 7.09) respectively in favour of social activities.
Carer interventions
Two studies (77 participants) investigated carer interventions. Only one reported objective measures, describing 108 min more nocturnal sleep time, 8.4% greater sleep efficiency and 24 min less total wake time.
Multimodal interventions
Seven studies (589 participants) examined multimodal interventions. Nocturnal sleep time was 24 to 35.3 min greater with multimodal interventions than usual care, and sleep efficiency ranged from 3.8% lower to 4.8% greater. Two studies reported reduced total wake time with multimodal interventions, ranging from 7 to 36 min less. Nocturnal awakenings ranged from 4.7 fewer to 0.1 more. One study reported adverse events for the physical activity component of its multimodal intervention.
Other interventions
Three studies, respectively examining daytime sleep restriction, slow-stroke back massage and transcranial electrostimulation, reported no significant changes in sleep quality, nocturnal sleep time, sleep efficiency, sleep onset latency and objective sleep diaries.
Quality of evidence
Evidence from the studies in the review was deemed to be of low or very low certainty, for reasons including risk of bias, inconsistent results and imprecise results. The exception to this was evidence for sleep onset latency and adverse events in the studies comparing physical activities with usual care, which was rated as moderate certainty.
What can the evidence tell us?
The review authors acknowledged that clinical and methodological heterogeneity was a limitation of their review, which also meant that meta-analyses were not possible for most interventions (Box 2). Non-pharmacological interventions are highly varied, so some level of heterogeneity is likely unavoidable.
Types of heterogeneity
Heterogeneity refers to variation between studies, and various types can be considered.
Clinical heterogeneity suggests that trials are looking at different concepts (Fletcher Reference Fletcher2007). This might include different patient groups or the use of different interventions (Fletcher Reference Fletcher2007). In this review (Wilfling Reference Wilfling, Calo and Dichter2020), examples of clinical heterogeneity might include the range of interventions used – for example, in the light therapy group, studies varied in terms of the source of light, the brightness and colour of light, as well as duration of exposure.
Methodological heterogeneity includes differences attributable to variation in study design and how the study was completed (Gagnier Reference Gagnier, Moher and Boon2012). Examples evident in the studies used in this review could include that one of the light therapy studies used outcome measures assessed by unmasked (‘unblinded’) assessors, whereas several of the other studies used masked assessors – this difference in methods has introduced a difference in chance of bias between the two studies, which may yield differing results.
Statistical heterogeneity refers to when similar studies show variation in their results (Gagnier Reference Gagnier, Moher and Boon2012). This may be a consequence of clinical or methodological heterogeneity (Gagnier Reference Gagnier, Moher and Boon2012), and is usually only apparent after studies have been analysed, as separate statistical tests can be used to quantify the degree and significance of variation present (Fletcher Reference Fletcher2007).
It is important to consider heterogeneity because it informs how different study results can be compared and whether they are sufficiently similar for any comparison to be meaningful. This becomes particularly relevant when trying to combine the results of multiple studies as part of a meta-analysis, as this is possible only if the studies are sufficiently similar. In this review, we noted that this was not the case for most of the intervention groupings used.
The authors made a logical attempt to group interventions into similar categories. Although this is helpful in making the evidence more straightforward to understand, it resulted in quite different interventions or methods being grouped together. This may have contributed to the inconsistent findings, as even in the seemingly similar group of light therapy, the regimes varied considerably, from 400 to 12 000 lux per day.
The inclusion of a large number of studies, involving 1335 participants, is a strength of this review, but the quality of the studies has largely limited the possible conclusions. Positively, the review authors attempted to identify grey literature, ongoing and unpublished work. The majority of the studies took place in nursing homes and in the USA. Findings may therefore be less applicable to other settings.
The review’s broad inclusion criteria is a strength in terms of clinical applicability; however, this introduces clinical heterogeneity. The lack of restrictions on dementia type is a potential strength as it reflects the breath of presentations in clinical practice. However, the prevalence of sleep problems varies between dementia types, and the proposed underlying mechanisms of sleep difficulties may also differ between types (Cipriani Reference Cipriani, Lucetti and Danti2015; McCleery Reference McCleery and Sharpley2020). Given this, the effectiveness of interventions may vary between dementia types, making it difficult for any intervention to demonstrate consistently convincing findings. The use of a MMSE score less than 24 as an inclusion criterion is a further limitation, as this potentially introduces participants with cognitive impairment for reasons other than dementia (Räihä Reference Räihä, Isoaho and Ojanlatva2001; Paterniti Reference Paterniti, Verdier-Taillefer and Dufouil2002; Mitchell Reference Mitchell, Shukla and Ajumal2014). Such individuals might be expected to have a different response to any interventions for sleep disturbance, and were not the intended participants.
In some studies, the sleep parameters reported for the control group are close to normal values. Mean nocturnal sleep time in those over 65 years of age is around 486 min (Li Reference Li, Vitiello and Gooneratne2018). The nocturnal sleep time in the control groups for two of the light therapy studies was 512 and 496 min. Similarly, the mean nocturnal sleep time in the usual care group in the analysis for carer interventions was 468 min. This draws into question the usefulness of these groups as comparators.
What do these outcomes mean for patients?
This Cochrane Review reports a range of primary outcomes; however, the clinical relevance of these is unclear. Positively, the review included patient and public involvement (PPI) (Box 3), in the form of referencing six carer interviews completed as part of another Cochrane Review in the same field (McCleery Reference McCleery and Sharpley2020). On this basis, duration of consolidated sleep was added as a primary outcome. Unfortunately, none of the available studies reported on this. The use of objective sleep-related outcomes as a primary outcome was not clearly justified in the methods. The authors later describe polysomnography and actigraphy as being considered gold standard for assessing sleep-related outcomes, although of the two pieces of evidence cited, one involved adolescents and adults, and so may be less relevant to this population, and the other dates from 1997. They do not mention more recent evidence, such as a systematic review (Webster Reference Webster, Costafreda Gonzalez and Stringer2020) which suggested that actigraphy did not offer any specific benefit in assessing sleep in those with dementia in care homes, compared with proxy measures. A recent RCT raised issues with acceptability of the actigraphy devices, as well as discrepancies between actigraphy data and observed sleep, in people with dementia (Rapaport Reference Rapaport, Amador and Adeleke2024).
Patient and public involvement
Patient and public involvement (PPI) refers to the involvement of patients, or those with relevant experience, in the design, completion or dissemination of a study.
PPI is regarded as being crucial in ensuring that healthcare research is relevant to patients and considers patients’ experiences and priorities (Arumugam Reference Arumugam, Phillips and Moore2023), and its use is reflected in national guidance (National Institute for Health and Care Excellence 2024).
The National Health Service (NHS) Health Research Authority (HRA) has outlined four best practice principles to ensure meaningful PPI in research (HRA 2025).
Involve the right people – this generally includes people with relevant lived experience. In this review (Wilfling Reference Wilfling, Calo and Dichter2020), it might include those with dementia and their carers, although the PPI for this review only reflected the experience of carers, and not those with dementia themselves.
Involve enough people – this means a sufficient number of people to provide with breath and depth required, which will vary considerably between studies. For this review, it might have been helpful to consider including those with experience of a range of dementia types, severities and clinical settings.
Involve those people enough – this refers to involving those with lived experience in as many parts of the study as is appropriate. The scope and opportunities for involving those with lived experience may be broader than it initially appears, and could include using PPI in determining outcome measures used (as was the case in this review), informing methods used, for example by advising on potential barriers and difficulties that people might experience during research, explaining the relevance and acceptability of research as part of funding or ethics applications, as well as assisting with developing plain language research summaries or contributing towards dissemination of results.
Describe how it helps – this includes outlining how the public are being involved in the project itself, which will indicate how a study is relevant to them. Research Ethics Committees use this information in their review of research applications.
Given that more subjective measures, such as patient distress, carer burnout and earlier need for residential care, are potential consequences of sleep problems in dementia, there may be a greater role for carer and patient involvement to inform the clinical importance of sleep-related outcomes in this group.
Conclusions
This review suggests potential benefits of physical activity, social activities and carer interventions for sleep problems in dementia, but this remains inconclusive, owing to the heterogeneity and quality of available RCTs.
The review prioritised objective measures of sleep, which may not necessarily reflect outcomes most relevant to patients and carers. More integrated involvement of these groups could be helpful to inform relevant outcomes in future research.
Author contributions
F.C. drafted the initial manuscript, both N.B. and T.B. reviewed and made further edits to the manuscript. All authors have reviewed and approved the final version to be submitted.
Funding
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Declaration of interest
None.
eLetters
No eLetters have been published for this article.