Sleep disturbances are highly prevalent among patients with dementia and are associated with significant negative outcomes, including increased caregiver burden, reduced quality of life, behavioral disturbances, and a higher risk of institutionalization (Petrovsky, 2018; Bombois, 2010). Sleep disturbances can also had a negative impact on cognition. Management of these sleep disturbances is challenging and often includes both non-pharmacological and pharmacological interventions (Boeve, 2008; Rose, 2010).

 

In current clinical practice, there is considerable variation in how these treatment options are applied. This variability is due to limited and sometimes conflicting evidence on the effectiveness and safety of both non-pharmacological strategies (e.g., sleep hygiene, light therapy, behavioral interventions) and pharmacological agents (e.g., sedative-hypnotics, antipsychotics, melatonin). In addition, the implementation of non-pharmacological approaches can be resource-intensive and inconsistently applied (Rose, 2010), while the risk of adverse effects -such as increased fall risk, excessive sedation, and potential cognitive worsening- complicates pharmacological decision-making (Bombois, 2010; Boeve, 2008).

 

The underlying issue prompting this key question is the lack of clear, evidence-based guidance to support consistent and effective management of sleep disorders in people with dementia. Addressing this gap is necessary to reduce practice variation, improve patient outcomes, and guide clinicians in choosing appropriate, safe, and acceptable interventions-both pharmacological and non-pharmacological. Specific sleep disorders in people with dementia are common; insomnia, circadian rhythm disorders, Restless Legs Syndrome, sleep apnea, REM- sleep behavior disorder. The aim of this search focuses on objective sleep disturbances in people with dementia. We do not discriminate between the different sleep disorders. The sleep disturbances will reflect mainly insomnia and/or circadian rhythm disorders. For management of sleep disturbances, it is important to discriminate between the different sleep disorders. The management of the different sleep disorders will not be discussed in the search. Also, mild cognitive impairment (MCI) is not addressed in this module. For people without cognitive impairment, cognitive behavioral therapy for insomnia is the preferred treatment. Because most patients with MCI are still capable to learn new skills, cognitive behavioral therapy can be used in these patients.

Summary of Findings

Subquestion 1.1: Non-pharmacological treatments

1.1.1 Light-therapy

Population: Older adults (usually ≥65 years) with any form of dementia (e.g., Alzheimer’s, Lewy body, vascular, frontotemporal) experiencing sleep disturbances or disorders in the hospital setting  (both outpatient and hospital admission).

Intervention: Bright-light exposure delivered via light boxes or ceiling fixtures, typically 2,000–10,000 lux, once or twice daily for 30–120 minutes, over 4 to 10 weeks.

Comparator: Usual care, dim light, or sham light (<300 lux).

Setting/measurement: Sleep assessed with actigraphy or polysomnography (TST, sleep efficiency, wake time, awakenings, sleep latency) and subjective scales (PSQI or Sleep Disorders Inventory) in subsets.

Outcome	Study results and measurements	Absolute effect estimates	Certainty of the evidence (Quality of evidence)	Summary
Sleep quality (objective or subjective)	Objective sleep assessed via actigraphy/polysomnography in 6 RCTs (n = 302 unique participants) (Fontana Gasio 2003; Dowling 2005; McCurry 2011; Nowak 2008; Figueiro 2019; Ancoli-Israel 2003). Outcomes: total sleep time (TST), sleep efficiency (SE), wake after sleep onset (WASO), nocturnal awakenings, sleep onset latency (SOL). Subjective sleep in 2 RCTs (n = 108 participants; Figueiro 2019 [PSQI], McCurry 2011 [SDI]).	TST MDs ranged from −33 to +110 min; pooled effect ≈ +18 min. SE MDs −4.5 to +16.6 % (pooled ≈ +1.5 %). WASO MD ≈ −8 min. SOL MD ≈ −1.8 min (95 % CI −5.4 to 1.9). Subjective sleep quality (SMD ≈ −0.25, 95 % CI −0.6 to 0.1).	Very low, downgraded for risk of bias, imprecision, and inconsistency.	We are uncertain whether light therapy improves or worsens sleep quality.
Daytime functioning / agitation	Four RCTs (n = 218; Ancoli-Israel 2003; Dowling 2005; Fontana Gasio 2003; McCurry 2011) assessed daytime behaviour using the CMAI, ABRS, or NPI-NH.	No numerical effect estimates were reported. Effects were small and inconsistent across studies, with some suggesting slight improvement and others deterioration.	No grade	-
Adverse events (safety)	Reported qualitatively in 6 RCTs; no serious or treatment-related adverse events reported.	–	No grade	-
Quality of life	Not reported	-	No grade	-
Caregiver burden	Not reported	–	No grade	-

_{CMAI: Cohen-Mansfield Agitation Inventory;  NPI - NH: Neuropsychiatric Inventory Home Version; SMD: Standardized mean difference; MD: Mean difference}

 

1.1.2 Physical activity

Population: Older adults (usually ≥65 years) with any form of dementia (e.g., Alzheimer’s, Lewy body, vascular, frontotemporal) experiencing sleep disturbances or disorders in the hospital setting  (both outpatient and hospital admission).

Intervention: Structured physical-activity programmes (e.g., walking, chair exercises) 3–7 sessions per week, over 4 to 16 weeks.

Comparator: Usual care or minimal social contact (e.g., conversation only).

Outcome	Study results and measurements	Absolute effect estimates	Certainty of the evidence (Quality of evidence)	Summary
Sleep quality (objective or subjective)	Actigraphy in 2 RCTs (n = 167; McCurry 2011: 32 I / 33 C; Richards 2011: 55 I / 47 C). Outcomes: TST, SE, WASO, nocturnal awakenings, SOL. One additional trial (Chan 2016; n = 91) reported qualitative findings only. Subjective sleep quality assessed in McCurry 2011 using a caregiver-rated Sleep Disorders Inventory (SDI) (n = 65; 32 intervention / 33 control).	Objective sleep quality across studies: TST ≈ 10–40 min longer with activity; SE +2 to +5 %; WASO ≈ 30 min lower; nocturnal awakenings ≈ 3 fewer per night; SOL no clear change.   Subjective sleep quality showed minimal between-group difference (Δ −0.6 vs −0.2 points on the SDI; lower = better sleep).	Very low, downgraded for risk of bias (unblinded behavioural interventions) and imprecision (small samples).	We are uncertain whether carer-focused interventions influence physical or social activity.
Daytime functioning / agitation	Reported qualitatively in McCurry 2011 (n = 65) using daytime alertness ratings; no standardized scales used. No numerical estimates reported.	–	No grade	-
Adverse events (safety)	Explicitly monitored in both trials (n = 167; McCurry 2011 and Richards 2011). Richards 2011 reported three transient, non-serious events (one chest-pain episode, one musculoskeletal pain, one ECG anomaly). No withdrawals or serious adverse events were reported.	–	No grade	-
Quality of life	Not assessed in any included study.	–	No grade	-
Caregiver burden	Not assessed in any included study.	–	No grade	-

 _{SMD: Standardized mean difference; MD: Mean difference}

 

1.1.3 Social activity

Population: Older adults (usually ≥65 years) with any form of dementia (e.g., Alzheimer’s, Lewy body, vascular, frontotemporal) experiencing sleep disturbances or disorders in the hospital setting  (both outpatient and hospital admission).

Intervention: Individualized, one-to-one social activities (e.g., conversation, music, crafts, letter writing) for 1–2 hours daily, over 3 to 7 weeks.

Comparator: Usual care or minimal social contact (e.g., conversation only).

Outcome	Study results and measurements	Absolute effect estimates	Certainty of the evidence (Quality of evidence)	Summary
Sleep quality (objective or subjective)	Measured by actigraphy (total nocturnal sleep time [TST], sleep efficiency [SE]) in 2 RCTs (n = 236; Richards 2005: 71 I / 68 C; Richards 2011: 50 I / 47 C).	Across studies, TST was approximately 13–29 minutes longer and SE about 1–6 percentage points higher in social-activity groups compared with usual care. Other actigraphic parameters (WASO, awakenings, SOL) were not reported in a comparable format.	Very low, downgraded for risk of bias (unblinded behavioural interventions) and imprecision (small samples).	We are uncertain whether carer-focused interventions influence physical or social activity.
Daytime functioning / agitation	Not assessed.	–	No grade	No evidence available.
Adverse events (safety)	Explicitly monitored in both trials (n = 167; McCurry 2011 and Richards 2011). Richards 2011 reported three transient, non-serious events (one chest-pain episode, one musculoskeletal pain, one ECG anomaly). No withdrawals or serious adverse events were reported.	–	No grade	-
Quality of life	Not assessed.	–	No grade	No evidence available.
Caregiver burden	Not assessed.	–	No grade	No evidence available.

 _{SMD: Standardized mean difference; MD: Mean difference}

 

1.1.4 Carer-focused

Population: Older adults (usually ≥65 years) with any form of dementia (e.g., Alzheimer’s, Lewy body, vascular, frontotemporal) experiencing sleep disturbances or disorders in nursing home or residential care settings.

Intervention: Education or training programmes for caregivers focused on sleep-hygiene principles, behavioural management, and environmental optimisation. Duration: typically 4–12 weeks, delivered individually or in small groups.

Comparator: Usual care or provision of written sleep information only.

Outcome	Study results and measurements	Absolute effect estimates	Certainty of the evidence (Quality of evidence)	Summary
Sleep quality (objective or subjective)	Measured by actigraphy: total sleep time (TST), sleep efficiency (SE), and total wake time at night (WASO). Based on 2 RCTs (n = 77; McCurry 2012: 23 I / 10 C; Gattinger 2017: 22 I / 22 C).	Across studies, small and inconsistent improvements were seen in favour of the intervention: TST and SE slightly higher (≈ +5 – 10 %), WASO modestly lower (≈ 20–40 min). No pooled estimate calculated due to heterogeneity and small sample size.	Very low, downgraded for risk of bias (unblinded caregivers) and imprecision (small n).	We are uncertain whether carer-focused interventions influence sleep quality..
Daytime functioning / agitation	Not assessed.	–	No grade	No evidence available.
Adverse events (safety)	Two RCTs (McCurry 2011; Wilfling 2023) explicitely monitered adverse events in 77 participants.	No unexpected or serious events occurred.	No grade	-
Quality of life	Not assessed.	–	No grade	No evidence available.
Caregiver burden	Not assessed.	–	No grade	No evidence available.

 _{SMD: Standardized mean difference; MD: Mean difference}

 

1.1.5 Multimodal

Population: Older adults (usually ≥65 years) with any form of dementia (e.g., Alzheimer’s, Lewy body, vascular, frontotemporal) experiencing sleep disturbances or disorders in the hospital setting  (both outpatient and hospital admission).

Intervention: Combined non-pharmacological programmes incorporating two or more elements (e.g., light exposure + physical or social activity + sleep-hygiene education + environmental optimisation). Duration 4–16 weeks.

Comparator: Usual care or attention control without structured multimodal content.

Outcome	Study results and measurements	Absolute effect estimates	Certainty of the evidence (Quality of evidence)	Summary
Sleep quality (objective or subjective)	Objective sleep measured by actigraphy or polysomnography (TST, SE, WASO, nocturnal awakenings) across 7 RCTs (Alessi 1999; Alessi 2005; Li 2009; McCurry 2005; McCurry 2011; Richards 2011; Schnelle 1999; ≈ 590 participants). Several studies (e.g. Alessi 2005; Schnelle 1999) reported only group-level summaries or incomplete denominators, precluding reconstruction of exact sample sizes per outcome. Subjective sleep quality was assessed in two RCTs (Li 2009; McCurry 2011).	Across studies: TST generally 20–40 minutes longer, SE ~2–5 % higher, and WASO 10–40 minutes shorter in intervention groups; findings small and inconsistent. No pooled estimate presented due to heterogeneity and incomplete denominators.	Very low, downgraded for risk of bias (unblinded), imprecision (small samples, wide CIs), and inconsistency (heterogeneous interventions).	We are very uncertain about the effect of multimodal programmes on sleep quality.
Daytime functioning / agitation	Observed agitation measured in one study (Alessi 1999; n = 39 [19 I / 20 C]) using direct observation logs.	Fewer agitation episodes in the intervention group (mean 9.4 → 7.3 vs 5.9 → 14.7 in controls).	Very low, downgraded for risk of bias (unblinded) and imprecision (small sample, wide CIs).	We are very uncertain about the effect of multimodal programmes on daytime functioning.
Adverse events (safety)	Explicitly reported in one RCT (Richards 2011; n = 88 [41 I / 47 C]).	Three transient, non-serious events (one chest pain – no MI; one musculoskeletal complaint; one ECG anomaly). No serious or treatment-related events in other trials mentioning safety monitoring.	Very low, downgraded for risk of bias (unblinded) and imprecision (small sample, wide CIs).	We are very uncertain about the relationship between multimodal programmes and adverse events.
Quality of life	Not assessed.	–	No grade	No evidence available.
Caregiver burden	Not assessed.	–	No grade	No evidence available.

_{SMD: Standardized mean difference; MD: Mean difference}

 

1.1.6 Short-term

Population: Older adults (usually ≥65 years) with any form of dementia (e.g., Alzheimer’s, Lewy body, vascular, frontotemporal) experiencing sleep disturbances or disorders in the hospital setting  (both outpatient and hospital admission).

Intervention: Brief or single-component interventions such as massage therapy, transcranial electrical stimulation, or short sleep-restriction protocols. Duration: 1 to 4 weeks.

Comparator: Usual care or sham intervention.

Outcome	Study results and measurements	Absolute effect estimates	Certainty of the evidence (Quality of evidence)	Summary
Sleep quality (objective or subjective)	Ancoli-Israel (2003) (n = 38; 19 intervention / 19 control) assessed daytime sleep restriction, during which participants were kept awake for six hours daily (09:00–12:00 and 14:00–17:00) to discourage daytime napping. Sleep–wake patterns were recorded with actigraphy over seven days.   Harris (2012) (n = 29; 15 intervention / 14 control) tested slow-stroke back massage, applied for three minutes at bedtime over two consecutive nights by a trained geriatric nurse. Sleep quality was assessed through caregiver observation rather than actigraphy or validated scales.   Hozumi (1996) (n = 20; 10 intervention / 10 control) evaluated transcranial electrostimulation using a HESS-100 device (6–8 V, 20 min/day for two weeks). Outcomes labelled “sleep disorder improvement” were rated by clinicians on a non-validated categorical scale.	Ancoli-Israel (2003) reported daytime wakefulness increased from 65 % to 68 % in the intervention group, but total nocturnal sleep time and circadian rhythm parameters showed no meaningful differences between groups.   Harris (2012) noted residents “settled more easily” after massage, but no quantitative sleep data were reported.   Hozumi (1996) found 8 / 10 participants improved vs 2 / 10 in the control group, but the scoring method were not reported.	No grade	Evidence was insufficient and heterogeneous; no reliable quantitative estimates available.
Daytime functioning / agitation	Not measured in any included study.	–	No grade	No evidence available.
Adverse events (safety)	Safety explicitly reported in all three studies. (Ancoli-Israel 2003; Harris 2012; Hozumi 1996).	No serious or treatment-related adverse events were reported in any of these studies.	No grade	-
Quality of life	Not measured in any included study.	–	No grade	No evidence available.
Caregiver burden	Not measured in any included study.	–	No grade	No evidence available.

_{SMD: Standardized mean difference; MD: Mean difference}

 

Subquestion 1.2: Pharmacological treatment

1.2.1. Melatonin

Population: Older adults (usually ≥65 years) with any form of dementia (e.g., Alzheimer’s, Lewy body, vascular, frontotemporal) experiencing sleep disturbances or disorders in the hospital setting  (both outpatient and hospital admission).

Intervention: Melatonin 2–10 mg daily (oral), administered at bedtime. Treatment duration: 2 to 24 weeks across trials.

Comparator: Placebo

Outcome	Study results and measurements	Absolute effect estimates	Certainty of the evidence (Quality of evidence)	Summary
Sleep quality (objective or subjective)	Objective Measured by actigraphy (total nocturnal sleep time [TNST], sleep efficiency (SE), wake after sleep onset (WASO), number of nocturnal awakenings, mean duration of nocturnal sleep bouts, and ratio of daytime to night-time sleep). Data from 2 double-blind RCTs (n = 184; Singer 2003; Dowling 2008). Subjective Subjective sleep outcomes were reported using 5-point caregiver-rated sleep diaries (Singer 2003) or the Pittsburgh Sleep Quality Index (PSQI) (Morales-Delgado 2018).	TNST: –32.0 to +8.4 min (95% CI –104.9 to +49.5); SE: –0.01 (95% CI –0.06 to +0.04); WASO: +9.1 min (95% CI –7.5 to +25.7); Nocturnal awakenings: +6.0 (95% CI –2.7 to +14.6); Mean sleep-bout duration: –2.0 (95% CI –8.3 to +4.3); Daytime/night-time ratio: –0.11 to –0.25 (95% CI –0.71 to +0.37).   Subjective SMD –0.21 (95% CI –0.71 to 0.29) (Singer 2003). PSQI improved slightly in both groups; no numeric data provided (Morales-Delgado 2018).	Very low — downgraded for serious imprecision (small sample size, wide confidence intervals), risk of bias (unclear methods), and indirectness (heterogeneous formulations and co-interventions).	The effect on sleep quality is very uncertain.
Daytime functioning / agitation	Performance of activities of daily living (ADL): Alzheimer’s Disease Cooperative Study–ADL (Singer 2003) and Lawton IADL (Morales-Delgado 2018; Wade 2014).	SMD –0.04 (95% CI –0.34 to 0.26); individual study estimates –0.30 (Morales-Delgado 2018) to +0.08 (Singer 2003).	Very low — downgraded for imprecision (small, heterogeneous studies) and unclear risk of bias in behavioural assessments.	The effect on daytime functioning/agitation is very uncertain.
Adverse events (safety)	Three studies reported adverse events (Singer 2003; Morales-Delgado 2018; Wade 2014). Typical events: somnolence, dizziness, headache (<10%). No withdrawals due to AEs; no excess falls/cognitive decline.	Event rates similar to placebo across trials; no serious AEs attributed to melatonin.	Low — downgraded for imprecision (few events, limited sample) and incomplete reporting across studies.	Melatonin may have a safety profile similar to placebo in the short term.
Quality of life	Wade 2014: QoL-AD measured but not reported; authors stated no significant difference between groups.	No numerical data.	No grade	No evidence available.
Caregiver burden	Morales-Delgado 2018 (IR 5 mg, 8 weeks; n = 31): Zarit Burden Interview MD –6.20 (95% CI –19.81 to +7.41).	MD –6.20 (95% CI –19.81 to +7.41).	Very low — downgraded for very serious imprecision (single small study) and risk of bias.	The effect on caregiver burden is very uncertain.

_{SMD: Standardized mean difference; MD: Mean difference}

 

1.2.2. Trazodone

Population: Older adults (usually ≥65 years) with any form of dementia (e.g., Alzheimer’s, Lewy body, vascular, frontotemporal) experiencing sleep disturbances or disorders in the hospital setting  (both outpatient and hospital admission).

Intervention: Trazodone 50 mg daily (oral), administered at bedtime for 2 weeks.

Comparator: Placebo

Outcome	Study results and measurements	Absolute effect estimates	Certainty of the evidence (Quality of evidence)	Summary
Sleep quality (objective or subjective)	Actigraphy (TST, SE, WASO, sleep latency). 1 RCT, n=30, 2 weeks (Camargos 2014; 50 mg nightly).	TST +42.5 min (95% CI +0.9 to +84.1); SE +8.5% (95% CI 1.9 to 15.1); WASO −20.41 min (95% CI −60.4 to 19.6); no effect on sleep latency. Effects small to modest.	Very low — downgraded for very serious imprecision (single small study, n=30; wide confidence intervals) and risk of bias (short duration, unclear allocation concealment).	The effect on objective sleep parameters over 2 weeks is very uncertain.
Daytime functioning / agitation	Actigraphic daytime counts and caregiver diaries (same trial).	No clinically relevant between-group differences.	Very low — downgraded for very serious imprecision (single small study) and potential bias in subjective caregiver-rated outcomes	The effect on daytime functioning/agitation is very uncertain.
Adverse events (safety)	Camargos 2014 adverse events (AE)s mild/infrequent: somnolence (2 participants), dry mouth (1). No SAEs or withdrawals due to AEs; no falls/cognitive worsening.	4/15 (trazodone) vs 6/15 (placebo); RR 0.67 (95% CI 0.23–1.89).	Low — downgraded for imprecision (small total, few events) and limited generalizability (short follow-up, older population).	Short-term trazodone may have a safety profile similar to placebo.
Quality of life	Not assessed.	–	No grade	No evidence available.
Caregiver burden	Not reported.	–	No grade	No evidence available.

_{SMD: Standardized mean difference; MD: Mean difference}

 

1.2.3. Ramelteon

Not enough information to generate summary of findings table.

 

1.2.4. Orexin receptor antagonists

Population: Older adults (usually ≥65 years) with any form of dementia (e.g., Alzheimer’s, Lewy body, vascular, frontotemporal) experiencing sleep disturbances or disorders in the hospital setting  (both outpatient and hospital admission).

Intervention: Orexin receptor antagonists (suvorexant 10–20 mg/day; lemborexant 2.5–15 mg/day), administered once daily at bedtime for up to 4 weeks.

Comparator: Placebo

Outcome	Study results and measurements	Absolute effect estimates	Certainty of the evidence (Quality of evidence)	Summary
Sleep quality (objective or subjective)	Polysomnography and actigraphy (TST, SE, WASO, sleep latency) plus sleep diaries/caregiver questionnaires. 2 double-blind RCTs, Herring 2020 (suvorexant 10–20 mg, 4 w; n≈285) and NCT03001557 (lemborexant 2.5–15 mg, 4 w; n=63).	Objective TNST (4.1): MD 28.20 min (95% CI 11.10 to 45.30; n = 274). WASO (4.4): MD −15.70 min (95% CI −28.10 to −3.30; n = 274). Sleep efficiency (4.3): MD 4.26% (95% CI 1.26 to 7.26; n = 312). Nocturnal awakenings (4.5): MD 0.00 (95% CI −0.50 to 0.50; n = 274). Sleep latency (4.6): MD −12.10 min (95% CI −25.90 to 1.70; n = 274). Mean duration sleep bouts (4.2): MD −2.42 min (95% CI −5.53 to 0.70; n = 38).   Subjective SDI (4.7): MD −0.09 (95% CI −0.19 to 0.01; n = 312)	Low — downgraded for indirectness (different drugs and measurement methods, short duration) and inconsistency (heterogeneity in sleep efficiency outcomes).	Orexin antagonists may improve objective sleep (TST and SE) over 4 weeks.
Daytime functioning / agitation	Not reported in either trial.	–	No grade	No evidence available.
Adverse events (safety)	2 double-blind RCTs, Herring 2020 (suvorexant 10–20 mg, 4 w) and NCT03001557 (lemborexant 2.5–15 mg, 4 w). Common AEs mild–moderate: somnolence 6–8%, headache ~5%, fatigue 3–4%. 1 suvorexant participant (<1%) discontinued for daytime sleepiness; no excess falls/cognitive decline.	≥1 AE (5.1): RR 1.29 (95% CI 0.83 to 1.99; 2 studies, n=323)	Low — downgraded for imprecision (wide confidence interval including possible harm) and short duration of follow-up.	Orexin antagonists may have a safety profile similar to placebo over 4 weeks.
Quality of life	Not reported in either trial.	–	No grade	No evidence available.
Caregiver burden	Herring 2020 (suvorexant 10–20 mg, 4 w; n≈285)	carer distress MD −0.1 (95% CI −0.2 to 0.1; n = 274)	Very low — downgraded for indirectness (carer distress ≠ caregiver burden), imprecision (single small study), and short duration.	The effect on caregiver burden is very uncertain.

_{SMD: Standardized mean difference; MD: Mean difference}

 

Table 2. Characteristics of included studies

Study	Participants	Comparison	Follow-up	Outcome measures	Comments	Risk of bias (per outcome measure)*
McCleery and Sharpley (2020)	RCTs = 9, including five trials on melatonin, one on trazodone, one on ramelteon, and two on orexin antagonists.   Total number of participants = 649.   Mean ages ranged from 72 to 85 years.   Most participants were female (approximately 60–85%).   Studies included individuals with mild-to-moderate dementia (n = 3), moderate dementia (n = 2), and moderate-to-severe dementia (n = 4), primarily with Alzheimer’s disease.	Intervention: Melatonin (5 trials): – Daily dose range: 2–10 mg (2–2.5 mg SR; 5–10 mg IR) – Duration: 2 to 24 weeks   Trazodone (1 trial): – 50 mg once daily – Duration: 2 weeks   Ramelteon (1 trial): – 8 mg once daily – Duration: 8 weeks   Orexin receptor antagonists (2 trials): – Suvorexant 10–20 mg (individually titrated), 4 weeks – Lemborexant 2.5–15 mg, 4 weeks   Control: Placebo —nine double-blind, placebo-controlled RCTs	Follow-up periods across the nine included RCTs ranged from 2 to 24 weeks, depending on intervention type. – Melatonin studies assessed outcomes mainly at 2–10 weeks, with one trial extending to 24 weeks. – The Trazodone trial evaluated effects after 2 weeks. – Ramelteon was assessed after 8 weeks. – Orexin antagonist studies (suvorexant, lemborexant) measured sleep outcomes after 4 weeks of treatment. No long-term (>6 months) follow-up data were available.	Sleep quality (objective or subjective): 9 studies. • Objective: 7 used actigraphy/polysomnography (total sleep time, sleep efficiency, WASO). • Subjective: 4 used sleep diaries or caregiver sleep quality ratings. Definitions varied across drug classes.   Daytime functioning / agitation: 3 studies, using Epworth Sleepiness Scale or Neuropsychiatric Inventory (NPI).   Adverse events (falls, somnolence, dizziness, cognitive effects): 9 studies. Somnolence and dizziness most frequent; 3 studies monitored cognitive effects. Data from spontaneous adverse event reporting and clinical observation.   Quality of life: 1 study (QoL-AD scale).   Caregiver burden: 1 study (carer-rated sleep disturbance and burden questionnaire). Definitions and measurement tools varied between drug classes (melatonin, trazodone, orexin antagonists).	Detailed information on study characteristics and risk of bias (RoB) of the included studies can be found in McCleery (2020).   The only ramelteon trial (NCT00325728) was commercially sponsored by the manufacturer. No peer-reviewed publication or detailed study report was available; results were derived only from a summary posted on the sponsor’s website. The limited availability of methodological information, combined with commercial funding, raises concern about potential reporting bias and incomplete outcome data.	Sleep quality (objective or subjective): Moderate risk of bias – Most trials had adequate randomisation but unclear allocation concealment and incomplete outcome data; objective measures (actigraphy/polysomnography) reduce detection bias.   Daytime functioning / agitation: High risk of bias – Few studies, small samples, and outcome definitions not prespecified.   Adverse events: Low risk of bias – Adverse events consistently reported; blinding of assessors maintained.   Quality of life: High risk of bias – Single small study with unclear blinding and incomplete data.   Caregiver burden: High risk of bias – Single small study; self-reported data with high attrition.
Wilfling (2023)	RCTs = 19, covering a wide range of non-pharmacological approaches (e.g. light therapy, exercise, behavioural, sensory, and multicomponent interventions). Total number of participants = approximately 1300. Mean ages ranged from 74 to 88 years. Most participants were female (60–80%). Participants were nursing home residents and had mild (n ≈ 5), moderate (n ≈ 7), or moderate-to-severe dementia (n ≈ 7), diagnosed according to established clinical criteria (most commonly Alzheimer’s disease, but also mixed and vascular dementia).	Intervention: Various non-pharmacological interventions were tested:   – Bright-light therapy (6 studies) with assessment after 2–11 weeks or 15 days   – Physical activity (2 studies) and social activities (2 studies), evaluated after 7 weeks and 2 months (physical) and after 21 days and 7 weeks (social)   – Caregiver/carer-based interventions (3 studies), assessed at ≈ 6 months   – Multimodal programs (7 studies) combining e.g. light therapy, exercise and sleep-hygiene training – durations ranging from 32 days to 6 months   – Other: restriction of daytime napping (1 study; 5 nights), slow-stroke back massage (1 study; 2 weeks), transcranial electrical stimulation (1 study; 6 weeks)   Control: Usual care (≈16 studies) or no-treatment controls; some trials included both or used attention-control conditions (e.g. dim light or social contact). In light-therapy trials, “dim-light” or attention-control conditions were commonly used. Several studies employed alternative non-pharmacological comparators (e.g. social activity or exercise controls), and in a few cases pharmacological comparators were included. Most meta-analyses used usual care as the reference condition.	Follow-up durations varied widely between interventions, ranging from 5 nights to 6 months. – Light-therapy studies measured outcomes after 2 to 11 weeks or 15 days.   – Physical and social activity programs were evaluated after 3–8 weeks or up to 2 months.   – Carer-focused interventions were followed up at approximately 6 months.   – Multimodal interventions had assessments from 32 days to 6 months, depending on the component mix.   Short-term interventions such as napping restriction or massage reported outcomes within 2–5 weeks. No studies provided data beyond 6 months.	Sleep quality (objective or subjective): 29 studies. • Objective: 18 used actigraphy and 3 polysomnography (total sleep time, sleep efficiency, awakening frequency). • Subjective: 14 used sleep diaries or scales (e.g. PSQI or NPI sleep subscale). Definitions varied by intervention type.   Daytime functioning/agitation (BPSD, mood, alertness): 12 studies, using NPI, CMAI, CSDD, and sleepiness ratings; 5 measured daytime naps/alertness, 4 agitation, 3 mood.   Adverse events: 3 studies explicitly reported none; no standardized safety assessment tools used.   Quality of life: 5 studies (QoL-AD, DEMQOL, or well-being scales), with mixed self vs proxy ratings.   Caregiver burden: 3 studies (Zarit Burden Interview or similar). Definitions and measurement tools varied across intervention types (light therapy, activity, carer training, multimodal).	Detailed information on study characteristics and risk of bias (RoB) of the included studies can be found in Wilfling (2023).	Sleep quality (objective or subjective): Moderate risk of bias – Randomisation and blinding often unclear; actigraphy adds objectivity but many studies relied on caregiver ratings.   Daytime functioning / agitation (BPSD, mood, alertness): High risk of bias – Behavioural outcomes mostly unblinded, small samples, high attrition.   Adverse events: Low risk of bias – Reporting generally complete; no serious adverse events identified.   Quality of life: Moderate to high risk of bias – Inconsistent measurement instruments and frequent missing data.   Caregiver burden: High risk of bias – Few studies, self-reported measures, and limited blinding.

_{*For further details, see risk of bias table in the appendix}

 

Results

1. Subquestion – non-pharmacological treatment (Wilfling, 2023)

1.1 Light-therapy

Seven RCTs (Ancoli-Israel, 2003; Dowling, 2005; Fontana Gasio, 2003; McCurry, 2011; Nowak, 2008; Figueiro, 2019; Sloane, 2015) with a total of 363 unique participants evaluated bright-light exposure (morning, afternoon, evening, or combined; 2 500-13 000 lux) compared with dim-light or usual care for 2-16 weeks. For most studies, participant numbers per intervention group were approximately balanced (about 176 in light-therapy arms versus 126 in control conditions).

 

Exact allocation could not always be verified, as some trials (Ancoli-Israel, 2003) included multiple light-exposure arms (morning, evening, or combined) that were aggregated into a single “light vs control” comparison in the Cochrane review. Light-therapy protocols varied in timing and intensity, including exposure to bright white or blue-green light, blue-white ambient lighting installations, and dawn-dusk simulations.

 

Sleep quality (objective or subjective)

Sleep quality (objective)

Objective sleep outcomes were measured using actigraphy or polysomnography in six studies (Ancoli-Israel, 2003; Dowling, 2005; Fontana Gasio, 2003; McCurry, 2011; Nowak, 2008; Figueiro, 2019), assessing total nocturnal sleep time, sleep efficiency, total wake time at night, number of awakenings, and sleep onset latency.

 

Pooling of results was not possible due to heterogeneity in light intensity, exposure timing, and study duration. Results are displayed in figure 1.

 

_{Figure 1. The effect of light-therapy compared to usual care on sleep quality (objective)}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

Sleep quality (subjective)

Two studies assessed caregiver- or patient-rated sleep quality: Figueiro (2019) (n=41; 21 light vs 20 control) using the Pittsburgh Sleep Quality Index (PSQI), and McCurry (2011) (n=67; 34 light vs 33 control) using the Sleep Disorders Inventory (SDI). Results are displayed in figure 2. Pooling of results was not possible because only two studies reported data for this outcome, and due to heterogeneity in the studies.

 

_{Figure 2. The effect of light-therapy compared to usual care on sleep quality (subjective)}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

Daytime functioning/agitation

Three studies (Ancoli-Israel, 2003; Dowling, 2005; Fontana Gasio, 2003) assessed the outcome daytime functioning/agitation using the Cohen-Mansfield Agitation Inventory (CMAI), Agitated Behaviour Rating Scale (ABRS), or Neuropsychiatric Inventory (Nursing Home Version) (NPI[-NH]). No numerical effect estimates were reported. Findings were inconsistent and small in magnitude: one study described reduced agitation following light therapy (Ancoli-Israel 2003), another reported slightly higher agitation scores in the intervention group (Dowling 2005), and Fontana Gasio (2003) found no meaningful differences between groups.

 

Adverse events (safety)

Across all included studies (Ancoli-Israel, 2003; Dowling, 2005; Fontana Gasio, 2003; McCurry, 2011; Nowak, 2008; Figueiro, 2019), no adverse or serious adverse events were reported. None of the trials mentioned treatment-related discontinuations or ocular complaints.

 

Quality of life

No studies reported on quality of life measures.

 

Caregiver burden

No studies reported on caregiver burden.

 

1.2 Physical activity

Three RCTs (Chan, 2016; McCurry, 2011; Richards, 2011) with a total of 258 participants evaluated the effects of structured physical-activity programmes (walking, resistance training, or Tai Chi Qigong) over 7 weeks to 2 months on sleep quality.

Sleep quality (objective or subjective)

Sleep quality (objective)

Two RCTs (n = 167; McCurry 2011: n=65 [32 intervention/33 control]; Richards 2011: n=102 [55 intervention/47 control]) contributed quantitative data to the Cochrane analyses measured by actigraphy or polysomnography on total nocturnal sleep time, sleep efficiency, total wake time at night, number of nocturnal awakenings or sleep-onset latency.

Chan (2016) (n=91; 45 intervention/46 control) provided only qualitative information and did not report actigraphic sleep outcomes.

 

Results are displayed in figure 3. Pooling of results was not possible because only one or two studies per subgroup reported data for this outcome, and due to heterogeneity in the studies.

 

_{Figure 3. The effect of physical activity compared to usual care on sleep quality (objective)}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

For sleep latency, Richards (2011) found no changes after six months (no further information reported).

 

Sleep quality (subjective)

McCurry (2011) (n=65; 32 intervention/33 control) assessed caregiver-rated sleep quality and found minimal change between groups (mean change −0.6 vs −0.2; no meaningful difference).

 

Daytime functioning/agitation
Not formally assessed in the physical-activity subgroup. McCurry (2011) reported a slight improvement in daytime alertness, but no consistent changes in mood or agitation.

 

Adverse events (safety)
Across studies, adverse events were rare. Richards (2011) (n=102; 55 intervention/47 control) reported three transient, non-serious events: one episode of chest pain (no myocardial infarction), one case of back, hip and leg pain, and one ECG anomaly. No serious or treatment-related adverse events occurred in other studies.

 

Quality of life
No studies reported on quality of life measures.

 

Caregiver burden
No studies reported on caregiver burden.

 

1.3 Social activity

Two RCTs (Richards, 2005; Richards, 2011) compared individualized social-activity programs (n=121) to usual care (n=115) in 236 nursing home residents with dementia. Participants in the intervention group received tailored one-to-one social activities (e.g., conversation, music, crafts, letter writing) delivered by trained staff for 1–2 hours daily, over 3 to 7 weeks.

 

Sleep quality (objective or subjective)

Sleep quality (objective)
Both studies measured sleep using actigraphy and reported total nocturnal sleep time (TST) and sleep efficiency (SE). Other actigraphic parameters (wake after sleep onset, number of awakenings, or sleep-onset latency) were not reported in a comparable format and could not be analysed.

 

Results are displayed in figure 4. Pooling of results was not possible because only two studies per subgroup reported data for this outcome, and due to heterogeneity in the studies.

 

_{Figure 4. The effect of social activity compared to usual care on sleep quality (objective)}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

Sleep quality (subjective)
No studies reported on subjective sleep quality parameters.

 

Daytime functioning/agitation
No studies reported on daytime functioning/agitation.

 

Adverse events (safety)
No adverse or serious adverse events were reported.

 

Quality of life
No studies reported on quality of life.

 

Caregiver burden
No studies reported on caregiver burden.

 

1.4 Carer-focused

Two RCTs (McCurry, 2012; Gattinger, 2017) compared caregiver training and behavioural guidance (sleep-hygiene education, structured case-conference support, and implementation of individualised sleep plans) to usual care in 77 nursing home residents with dementia. Both trials involved short, structured education programmes (two to four sessions of 60-120 minutes) delivered to nurses or caregivers, focusing on recognising sleep-wake patterns and adapting care routines.

 

Sleep quality (objective or subjective)

Sleep quality (objective)
Both studies measured sleep using actigraphy, reporting total nocturnal sleep time (TST), sleep efficiency (SE), and total wake time at night (WASO). Sleep-onset latency, number of awakenings, and consolidated sleep were not reported in a comparable form.

 

Results are displayed in figure 5; no pooled estimate was calculated due to heterogeneity and small sample size.

 

_{Figure 5. The effect of carer-focused interventions compared to usual care on sleep quality (objective)}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

Sleep quality (subjective)
Neither study used validated subjective sleep-quality scales.

 

Daytime functioning/agitation

No studies reported on daytime functioning/agitation.

 

Adverse events (safety)

Both studies explicitly monitored safety outcomes (n=77). No unexpected or serious adverse events occurred; interventions were well tolerated.

 

Quality of life

No studies reported on quality of life.

 

Caregiver burden

No studies reported on caregiver burden.

 

1.5 Multimodal interventions

Seven RCTs (Alessi, 1999; Alessi, 2005; Li, 2009; McCurry, 2005; McCurry, 2011; Richards, 2011; Schnelle, 1999) evaluated multimodal programmes combining two or more intervention components such as light exposure, physical or social activity, sleep-hygiene education, environmental optimisation (e.g., noise or light reduction), and caregiver training.

Across studies, total sample size was approximately n=590, with group sizes roughly balanced between intervention and control arms. Exact numbers per comparison could not always be verified because several studies reported combined data across nested subgroups or incomplete participant allocation details.

Intervention duration ranged from 5 days to 14 weeks, and usual care typically consisted of standard nursing-home routines without structured sleep-related activities.

 

Sleep quality (objective or subjective)

Sleep quality (objective)
All seven studies measured sleep objectively using actigraphy or polysomnography, reporting total nocturnal sleep time (TST), sleep efficiency (SE), total wake time at night (WASO), and the number of nocturnal awakenings. Sleep-onset latency, consolidated sleep, and 24-hour sleep ratios were not consistently reported. Because several trials (Alessi, 2005; Schnelle, 1999) reported only group-level summaries or incomplete denominators, exact participant numbers contributing to each sleep outcome could not be reconstructed from the available data.

 

Results are displayed in figure 6. A pooled estimate was not calculated due to substantial heterogeneity in intervention composition, duration, and reporting formats.

 

_{Figure 6. The effect of multimodal interventions compared to usual care on sleep quality (objective)}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

Sleep quality (subjective)
Subjective sleep outcomes were inconsistently measured and incompletely reported, precluding pooled analysis.

 

Li (2009) evaluated perceived sleep quality using a locally developed, unvalidated sleep questionnaire. Scores improved substantially in the intervention group (mean score decreased from 13.6 at baseline to 2.2 post-intervention) compared with controls (13.0 to 4.9). However, the tool and scoring method were insufficiently described to allow quantitative comparison.

 

McCurry (2011) (n=66; 33 intervention/33 control) assessed subjective sleep using the Sleep Disorders Inventory (SDI), completed by caregivers. Mean SDI scores decreased slightly in both groups (intervention: 1.1 to 0.8; control: 0.8 to 0.5).

 

Daytime functioning/agitation

Alessi (1999) (n=39; 19 intervention/20 control) reported observed agitation episodes in the intervention group (mean number of observations with agitation: 9.4 [baseline] → 7.3 [follow-up]) compared to the control group (5.9 [baseline] → 14.7 [follow-up]).

 

Adverse events (safety)

Only Richards (2011) (n=88; 41 intervention/47 control) explicitly reported adverse events, describing three transient, non-serious cases: one episode of chest pain (no myocardial infarction), one musculoskeletal complaint (back, hip, and leg pain), and one electrocardiogram anomaly (multifocal premature ventricular contractions). No serious or treatment-related events were observed in other studies that mentioned safety monitoring.

 

Quality of life

No studies reported on quality of life.

 

Caregiver burden

No studies reported on caregiver burden.

 

1.6 Short-term interventions

Three small RCTs (Ancoli-Israel, 2003; Harris, 2012; Hozumi, 1996) evaluated brief, single-component interventions designed to improve sleep in residents with dementia.

 

Sleep quality (objective or subjective)

Ancoli-Israel (2003) (n=38; 19 intervention/19 control) assessed daytime sleep restriction, during which participants were kept awake for six hours daily (09:00-12:00 and 14:00-17:00) to discourage daytime napping. Sleep-wake patterns were recorded with actigraphy over seven days. Daytime wakefulness increased from 65% to 68% in the intervention group, but total nocturnal sleep time and circadian rhythm parameters showed no meaningful differences between groups.

 

Harris (2012) (n=29; 15 intervention/14 control) tested slow-stroke back massage, applied for three minutes at bedtime over two consecutive nights by a trained geriatric nurse. Sleep quality was assessed through caregiver observation rather than actigraphy or validated scales; staff noted that residents “settled more easily” after massage, but no quantitative sleep data were reported.

 

Hozumi (1996) (n=20; 10 intervention/10 control) evaluated transcranial electrostimulation using a HESS-100 device (6-8 V, 20 min/day for two weeks). Outcomes labelled “sleep disorder improvement” were rated by clinicians on a non-validated categorical scale; eight of ten participants in the intervention group were judged to have improved versus two of ten in the control group, though the scoring method and statistical significance were not reported.

 

Across these studies (total n≈87), interventions were brief and sample sizes small. Only one trial (Ancoli-Israel, 2003) used objective actigraphy, while others relied on qualitative or unvalidated measures.

 

Daytime functioning/agitation

No studies reported on daytime functioning/agitation.

 

Adverse events (safety)

No serious or treatment-related adverse events were reported in any of these studies.

 

Quality of life

No studies reported on quality of life.

 

Caregiver burden

No studies reported caregiver burden.

 

2. Subquestion – pharmacological treatment McCleery (2020)

2.1 Melatonin

Sleep quality (objective or subjective)

Four studies from McCleery (2020) evaluated melatonin in people with Alzheimer’s disease or mixed dementia who had a diagnosed sleep disturbance (Singer, 2003; Dowling, 2008; Morales-Delgado, 2018; Wade, 2014). Sleep was assessed objectively with actigraphy or polysomnography in two studies (Singer, 2003; Dowling, 2008). In addition, all studies also included caregiver-rated sleep diaries.

 

Sleep quality (objective), immediate-release melatonin

Results are displayed in figure 7. Pooling of results was not possible because only one or two studies reported data for this outcome, and due to heterogeneity in the studies.

 

_{Figure 7. The effect of immediate-release melatonin compared with placebo on sleep quality (objective)}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

Sleep quality (objective), slow-release melatonin

Results are displayed in figure 8. Pooling of results was not possible because only one study reported data for this outcome.

 

Figure 8. The effect of slow-release melatonin compared with placebo on sleep quality (objective)

Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval

 

Sleep quality (subjective)

Results are displayed in figure 9. Pooling of results was not possible because only two studies per subgroup reported data for this outcome, and due to heterogeneity in the studies.

 

_{Figure 9. The effect of immediate- and slow-release melatonin compared with placebo on sleep quality (subjective)}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

Daytime functioning/agitation

Three studies evaluated performance of activities of daily living (ADL) using the Alzheimer's Disease Cooperative Study-Activities of Daily Living (ADCS-ADL) scale (Singer, 2003) and Lawton's Instrumental Activities of Daily Living Scale (IADL) scale (Morales-Delgado, 2018; Wade, 2014).

 

Results are displayed in figure 10. Pooling of results was not possible because only two studies per subgroup reported data for this outcome, and due to heterogeneity in the studies.

 

_{Figure 10. The effect of immediate- and slow-release melatonin compared with placebo on daytime functioning}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

Adverse events (safety)

One study reported on the number of adverse events per person (Singer, 2003). Wade (2014) reported adverse event data for the whole trial population, so these data were not included. Morales-Delgado (2018) reported only that “the treatment was well-tolerated in all cases’’, and that no serious adverse events occurred during the trial, although they reported that one participant in the placebo group died.

 

Results are displayed in figure 11. Pooling of results was not possible because only one study reported data for this outcome.

 

_{Figure 11. The effect of immediate- and slow-release melatonin compared with placebo on adverse events}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

Quality of life

One study measured patients’ quality of life (Wade, 2014) but did not report the data. The authors stated that there was no significant difference between the treatment group.

 

Caregiver burden

One study (Singer, 2003) reported caregiver burden using the Zarit Burden Interview (Morales-Delgado, 2018). Results are displayed in figure 12. Pooling of results was not possible because only one study reported data for this outcome.

 

_{Figure 12. The effect of immediate-release melatonin compared with placebo on caregiver burden}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

2.2 Trazodone

One RCT (Camargos, 2014) including 30 participants with Alzheimer’s disease and sleep disturbance (15 receiving trazodone 50 mg once daily, 15 receiving placebo) compared trazodone with placebo for two weeks.

Sleep quality (objective or subjective)

Sleep quality (objective)

Objective sleep outcomes were measured using actigraphy, assessing total night sleep time (TNST), sleep efficiency, wake after sleep onset (WASO), number of night-time awakenings, daytime sleep and the number of daytime naps.

 

The mean difference (MD) for TNST was 42.46 minutes (95%CI 0.9 to 84.0); for sleep efficiency 8.53 percentage points (95%CI 1.9 to 15.1); for WASO -20.41 minutes (95%CI -60.4 to 19.6); for number of night-time awakenings -3.71 (95%CI -8.2 to 0.8); for daytime sleep 5.12 minutes (95% CI -28.2 to 38.4); and for daytime naps 0.84 (95%CI -2.6 to 5.4).

 

Sleep quality (subjective)

Caregiver-rated global sleep improvement was reported as “better” or “much better” for 10 of 15 participants in the trazodone group and 9 of 15 in the placebo group.

 

Daytime functioning/agitation

Daytime functioning was evaluated with the Katz Index of ADL. The MD for the Katz Index was 0.5 (95%CI -0.8 to 1.8).

 

Adverse events (safety)

Adverse events were reported by 4 of 15 participants in the trazodone group and 6 of 15 in the placebo group (RR: 0.67, 95%CI 0.23 to 1.89).

 

No serious adverse events, falls, or cognitive worsening were reported. Results are displayed in figure 13. Pooling of results was not possible because only one study reported data for this outcome.

 

_{Figure 13. The effect of trazodone compared with placebo on adverse events}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

Quality of life

Camargos (2014) did not report on quality of life.

 

Caregiver burden

Camargos (2014) did not report on caregiver burden.

 

2.3 Ramelteon

One phase 2 RCT (NCT00325728) including 74 participants with mild or moderate Alzheimer’s disease investigated ramelteon 8 mg versus placebo for eight weeks. The study was a multicentre, parallel-group study. The only available information was a clinical trial synopsis published by the sponsor.

 

Sleep quality (objective or subjective)

Sleep quality (objective)

Sleep outcomes were measured using actigraphy at weeks 1, 2, 4, 6, and 8. The primary outcome was TNST at week 1. The MD for TNST at week 1 was 18.2 minutes (95%CI -7.8 to 44.3). At week 1, daytime total sleep time was reported as 43.1 minutes higher (SE 16.19) in the ramelteon group; this difference was not reported at later time points. No between-group differences were reported for the percentage of participants with ≥30-minute increase in TNST, for wake after sleep onset, for sleep efficiency, for number of daytime naps, or for other nocturnal sleep parameters.

At weeks 1, 4, and 8, the ratio of daytime to night-time sleep was reported as higher in the ramelteon group, but no absolute values were provided.

 

Sleep quality (subjective)

NCT00325728 did not report on subjective sleep quality.

 

Daytime functioning/agitation

Exploratory outcomes were assessed at weeks 4 and 8. The only reported between-group difference was in the NPI disinhibition subscale at week 8 or early termination (MD: -0.9, SE 0.44). No other daytime or behavioural outcomes were reported.

 

Adverse events (safety)

Adverse events considered possibly, probably, or definitely related to study drug occurred in 21/74 participants (29.0% in the ramelteon group, 27.9% in the placebo group). Serious adverse events occurred in 4 participants in the placebo group and 0 participants in the ramelteon group. No deaths were reported.

 

Quality of life

NCT00325728 did not report on quality of life.

 

Caregiver burden

NCT00325728 did not report on caregiver burden.

 

2.4 Orexin receptor antagonists

Two RCTs investigated orexin antagonists in people with Alzheimer’s disease and sleep disturbance:

• Herring (2020): suvorexant 10-20 mg versus placebo, 4 weeks, n=285.
• NCT03001557: lemborexant 2.5-15 mg versus placebo, 4 weeks, n=63.

Both studies included participants with mild-to-moderate Alzheimer’s disease. Sleep outcomes were measured using polysomnography (Herring, 2020) and actigraphy (NCT03001557).

 

Sleep quality (objective or subjective)

Sleep quality (objective)

At four weeks, several sleep quality outcomes are available (see figure 14 and 15). Pooling of results was not possible because only one or two studies reported data for the outcome, and due to heterogeneity in the studies. At four weeks, several outcomes are available (see figure 14), including total nocturnal sleep time (TNST) was 28 minutes longer with suvorexant compared with placebo (mean difference [MD] 28.2 minutes, 95% CI 11.1 to 45.3; n = 274). The mean duration of sleep bouts was 2.4 minutes shorter with lemborexant (MD –2.42 minutes, 95% CI –5.53 to 0.70; n = 38, NCT03001557). No additional quantitative data on sleep-onset latency were available beyond week four.

 

_{Figure 14. The effect of orexin receptor antagonists compared with placebo on sleep quality}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

_{Figure 15. The effect of orexin receptor antagonists compared with placebo on sleep quality (mean duration of sleep bouts, change from baseline [minutes])}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

Sleep quality (subjective)
Results are displayed in figure 16. Pooling of results was not possible because only two studies reported data for this outcome, and due to heterogeneity in the studies.

 

_{Figure 16. The effect of orexin receptor antagonists compared with placebo on sleep quality (subjective)}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

Daytime functioning/agitation

Herring (2020) and NCT03001557 did not report on daytime activity, agitation, or activities of daily living.

 

Adverse events (safety)
Across both trials, adverse events were commonly reported but generally mild to moderate in intensity. The most frequent events were somnolence (6-8%), headache (5%), and fatigue (3-4%), with rates overall comparable to placebo. No increases in falls or cognitive decline were observed. One serious adverse event (a fall with fracture) occurred in the suvorexant group and was not considered related to study medication. No serious adverse events were reported in the lemborexant trial. One participant (<1%) in the suvorexant group discontinued treatment due to excessive daytime sleepiness.

 

Results are displayed in figure 17. Pooling of results was not possible because only two studies reported data for this outcome, and due to heterogeneity in the studies.

 

_{Figure 17. The effect of orexin receptor antagonists compared with placebo on adverse events}

_{Z: p-value of the pooled effect; df: degrees of freedom; I2: statistic heterogeneity; CI: confidence interval}

 

A systematic review of the literature was performed to answer the following question(s):

In older adults with dementia experiencing sleep disturbances, how do non-pharmacological and pharmacological interventions compare to placebo or usual care in improving sleep quality, behavioral symptoms, caregiver burden, and safety outcomes?

 

Table 1. PICO

Patients	Older adults (usually ≥65 years) with any form of dementia (e.g., Alzheimer’s, Lewy body, vascular, frontotemporal) experiencing sleep disturbances or disorders
Intervention-1	Any pharmacological (e.g., melatonin, sedative-hypnotics, antipsychotics) treatment for sleep problems
Intervention-2	Any non-pharmacological (e.g., sleep hygiene, light therapy, behavioral interventions) treatment for sleep problems
Control	Placebo, usual care, no intervention, or alternative interventions (pharmacological or non-pharmacological)
Outcomes	Sleep quality (objective or subjective), daytime functioning / agitation, adverse events (safety), quality of life, caregiver burden
Other selection criteria	Study design: systematic reviews

Relevant outcome measures

The guideline panel considered sleep quality (objective or subjective), as a critical outcome measure for decision making; and daytime functioning / agitation, adverse events (safety), quality of life, caregiver burden as an important outcome measure for decision making.

 

A priori, the guideline panel did not define the outcome measures listed above but used the definitions used in the studies.

 

In the literature there is no consensus about clinically relevant differences. The working group agreed on the following clinically relevant differences.

For sleep-specific parameters, a change of at least 30 minutes in total sleep time or 5 % in sleep efficiency was considered clinically relevant.

For behavioural and quality-of-life outcomes, a 0.5 SD improvement on validated scales (e.g. NPI, CMAI, QoL-AD, ZBI) was defined as the minimal clinically important difference.

 

For all other outcome measures, the guideline panel defined a minimal clinically (patient) important difference of a 25 % relative difference for dichotomous outcomes (0.8 ≤ RR ≥ 1.25) and 0.5 SD for continuous outcomes.

 

Search and select (Methods)

A systematic literature search was performed by a medical information specialist using the following bibliographic database: Embase.com. The database were searched from 2005 to 1 September 2025 for systematic reviews, RCTs and observational studies. A systematic search was completed using a combination of controlled vocabulary and natural language keywords. The overall search strategy was derived from the following primary search concepts: (1) dementia; (2) sleep disturbances or disorders. A total of 2361 records were imported for title/abstract screening. Ultimately, the title and abstract screening was restricted to systematic reviews (n=295). 

 

The search strategy was intentionally designed to be broad in scope in order to comprehensively address the research question. Rather than focusing on a single type of intervention, a wide range of treatment approaches was considered to ensure that all relevant evidence related to sleep disturbances in people with dementia was captured.

From this exploratory search, two recent high-quality Cochrane systematic reviews were identified that together comprehensively cover both pharmacological and non-pharmacological interventions for sleep disturbances in people with dementia: Wilfling (2023) and McCleery (2020).

 

These reviews adequately address the PICO of the current key question and provide a complete and up-to-date summary of the available evidence. The findings of these reviews are summarized in the literature summary.

1. Boeve BF. Update on the Diagnosis and Management of Sleep Disturbances in Dementia. Sleep Med Clin. 2008 Sep;3(3):347-360. doi: 10.1016/j.jsmc.2008.04.010. PMID: 21915174; PMCID: PMC3170725.
2. Bombois S, Derambure P, Pasquier F, Monaca C. Sleep disorders in aging and dementia. J Nutr Health Aging. 2010 Mar;14(3):212-7. doi: 10.1007/s12603-010-0052-7. PMID: 20191256.
3. Huo S, Cheng L, Li S, Xu F. Effects of eszopiclone on sleep quality and cognitive function in elderly patients with Alzheimer's disease and sleep disorder: A randomized controlled trial. Brain Behav. 2022 Feb;12(2):e2488. doi: 10.1002/brb3.2488. Epub 2022 Jan 18. PMID: 35041261; PMCID: PMC8865158.
4. McCleery J, Sharpley AL. Pharmacotherapies for sleep disturbances in dementia. Cochrane Database Syst Rev. 2020 Nov 15;11(11):CD009178. doi: 10.1002/14651858.CD009178.pub4. PMID: 33189083; PMCID: PMC8094738.
5. Petrovsky DV, McPhillips MV, Li J, Brody A, Caffeé L, Hodgson NA. Sleep disruption and quality of life in persons with dementia: A state-of-the-art review. Geriatr Nurs. 2018 Nov;39(6):640-645. doi: 10.1016/j.gerinurse.2018.04.014. Epub 2018 May 24. PMID: 29803606; PMCID: PMC6252149.
6. Rose KM, & Lorenz R. Sleep Disturbances in Dementia: What They Are and What To Do. Journal of Gerontological Nursing. 2010;36(5): 9–14. https://doi.org/10.3928/00989134-20100330-05 (Original work published May 1, 2010).
7. Wilfling D, Calo S, Dichter MN, Meyer G, Möhler R, Köpke S. Non-pharmacological interventions for sleep disturbances in people with dementia. Cochrane Database Syst Rev. 2023 Jan 3;1(1):CD011881. doi: 10.1002/14651858.CD011881.pub2. PMID: 36594432; PMCID: PMC9808594.