Light Intervention Effects on Circadian Activity Rhythm Parameters and Nighttime Sleep in Dementia Assessed by Wrist Actigraphy: A Systematic Review and Meta-Analysis

Abstract

Background and Objectives

Persons with dementia often show circadian rhythm disturbances and sleep problems. Timed light exposure seems to be a promising nonpharmacological treatment option. In this review, meta-analyses were run on light effects on circadian activity rhythm parameters in persons with dementia measured with wrist actimetry. Furthermore, we update a Cochrane review, published in 2014, on actigraphically measured light effects in nighttime sleep parameters in persons with dementia.

Research Design and Methods

Four electronic databases were searched for randomized controlled trials. Effects in meta-analyses were summarized by using mean differences and 95% confidence intervals. We followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines to assess the risk of bias and registered the review protocol (PROSPERO: CRD42020149001).

Results

Thirteen trials met inclusion criteria, and either utilized light therapy devices, ambient room lighting systems, or dawn–dusk interventions. Eleven of these studies were subjected to meta-analyses. They did not reveal significant light effects on circadian activity parameters: amplitude (p = .62; n = 313), acrophase (p = .34; n = 313), intradaily variability (p = .51; n = 354), and interdaily stability (p = .38; n = 354). Furthermore, no light effects were found on sleep parameters: total sleep duration (p = .53; n = 594), sleep efficiency (p = .63; n = 333), wake after sleep onset (p = .95; n = 212), and sleep onset latency (p = .26; n = 156). Subgroup analyses, pooling data from 3 studies including persons with Alzheimer’s dementia, also did not show light effects on circadian activity and sleep parameters. The overall risk of bias of included studies was high.

Discussion and Implications

There is insufficient evidence for actigraphically measured circadian light effects in persons with dementia. More high-quality research is needed to recommend the application of adjunctive light.

Actimetry, Dawn simulation, Light therapy, Rest–activity rhythm, Room light

Disruption of the circadian system is associated with poor physical and mental health in humans and increases the risks of diseases of the cardiovascular, metabolic, gastrointestinal, immune, and sleep–wake system (Allada & Bass, 2021; Roenneberg & Merrow, 2016). In addition, the onset of some diseases may be driven by the circadian clock, for example, the nighttime peak of asthma and the morning peak of myocardial infarction (Ruan et al., 2021). Furthermore, natural aging processes make the circadian system more susceptible to disturbances (Hood et al., 2017).

Alterations of circadian rhythmicity such as fragmented sleep–wake cycles (Bliwise, 1993) and reversed sleep–wake patterns are highly prevalent with the progression of dementia and may be partially attributed to a dysfunctional circadian system (Roth, 2012). To date, the management of sleep–wake problems in dementia includes pharmacological and nonpharmacological treatments. However, evidence for the efficacy of pharmacological interventions is still weak (McCleery et al., 2016). Preliminary findings (Dimitriou & Tsolaki, 2017) lent support to the hypothesis that bright light exposure may be a promising nonpharmacological treatment option to improve sleep and stabilize circadian activity profiles.

The light–dark cycle is the most important exogenous factor (Zeitgeber) for the regulation of the biological clock in humans (Roenneberg et al., 2013). The master clock of the circadian system is located in the suprachiasmatic nuclei (Golombek & Rosenstein, 2010) and processes photic input from the classical photoreceptor system and a small portion of retinal ganglion cells containing the photopigment melanopsin to diverse thalamic areas (Schmidt et al., 2011). These projections acutely modulate endocrine (Brainard et al., 2015), neurocognitive (Vandewalle et al., 2009), and alertness-related (Souman et al., 2018) parameters and change the amplitude of circadian parameters (Fonken & Nelson, 2014) or shift their phase (Stothard et al., 2017).

It is known that the intensity and spectral composition of light entering the eyes (Walmsley et al., 2015), light exposure history (Hebert et al., 2002), and the inner circadian clock time (Rüger et al., 2013) play a crucial role in the ability of light to affect the biological clock. An impaired visual system (e.g., lenses, retina, and optic nerve; van Someren et al., 1997) may reduce the photic input and thus the ability of light to synchronize and stabilize circadian rhythms with increasing age (Nylen et al., 2014). Furthermore, older adults and particularly persons with dementia are often exposed to reduced light levels in their living environments (Ancoli-Israel et al., 1997; Shochat et al., 2000) and tend to stay outdoors for only short periods (Duggan et al., 2008). Additionally, persons with dementia are often exposed to nighttime light when being cared for, which further disrupts nighttime sleep (Alessi et al., 2005).

The measurement of intervention effects in persons with dementia on nighttime sleep and circadian parameters is a practical challenge. On the one hand, polysomnographic recordings are hardly feasible and lead to high exclusion rates (for a review, see Camargos et al., 2013), whereas sleep logs and caregiver questionnaires often underestimate sleep times (McCurry et al., 2006). On the other hand, assessing the phase of the circadian pacemaker (Keijzer et al., 2014) by measuring nighttime saliva or blood melatonin levels or core body temperatures via a rectal sensor is hardly possible in this population. Body-worn accelerometers, so-called actimeters, enable long-term sleep–wake monitoring and have become an important objective assessment tool in sleep medicine (Smith et al., 2018) and chronobiology research (Wirz-Justice, 2007). Due to its nonobtrusiveness, actigraphy is increasingly used in interventional studies in patients with dementia.

Circadian activity rhythm parameters are derived from parametric and nonparametric analysis methods. The parametric analysis method fits a cosine curve to multiday activity data (Cornelissen, 2014) and is called cosinor analysis. The fitted curve is fully specified by three parameters: mesor (offset of the midline of the fitted cosine curve), amplitude (distance between mesor and the maximum of the fitted curve), and acrophase (time at which the first peak of the fitted curve occurs after a predefined reference time). Witting et al. (1990) and van Someren et al. (1996) co-developed a nonparametric circadian rhythm analysis method (NPCRA) for multiday activity data recorded from persons with dementia. NPCRA outcome parameters are mean activity level during the most inactive 5 h period (L5) and the most active 10 h period (M10). Furthermore, the similarity of daily activity cycles is derived from an interdaily stability (IS) measure that divides the variance of the mean 24 h activity profile by the variance of the multiday activity data. Finally, rest–activity fragmentation (i.e., repeated and extended inactive periods during the day and active periods during the night) is quantified by using an intradaily variability (IV) measure that is derived from the root mean square of the first (numerical) derivative of the mean hourly daily activity profile normalized by the variance of the multiday activity data.

The sleep scoring algorithm of Cole et al. (1992) is usually applied to actigraphy data recorded from persons with dementia. Common sleep outcome parameters measured by actimetry are total sleep time (number of minutes scored as sleep during the time in bed), sleep efficiency (the proportion of time scored as sleep during the time in bed), sleep onset latency (number of minutes from entering the bed till the first period of sleep), and the average length of nighttime awakenings (mean duration of the wake periods during the time in bed).

This systematic review summarizes circadian and sleep-related effects of adjunctive light interventions in persons with dementia from high-quality research (i.e., randomized controlled trials) measured by wrist actigraphy. For the first time, potential light effects on parametric and nonparametric circadian activity rhythm parameters are reviewed systematically and subjected to meta-analyses. Furthermore, meta-analyses for actigraphically measured light effects on sleep parameters, reported by Forbes et al. (2014) in a Cochrane review, are updated.

Method

To summarize results from studies with minimal risk of bias, this review compiles the effects of randomized controlled trials. Study participants were included with different types and specific stages of dementia. Included studies were approved by ethical committees and either implemented light therapy protocols (i.e., exposure to white light between 2,500 to 10,000 lux), room light interventions at any intensity, duration, and timing, or dawn–dusk simulation. Light intervention effects were measured with wrist actimetry and either reported as circadian activity rhythm or sleep parameters.

Light intervention studies, published in peer-reviewed English journals, were identified by searching the following four electronic databases: PUBMED, comprising biomedical literature from Medline, life science journals, and online books (https://www.ncbi.nlm.nih.gov/pubmed/); ALOIS, a trial register of dementia studies hosted by the Cochrane Dementia and Cognitive Improvement Group (www.medicine.ox.ac.uk/alois); Google Scholar, a web-search engine for scholarly literature across various publishing formats and disciplines (https://scholar.google.com); and SCOPUS, one of the largest database of peer-reviewed literature (https://www.scopus.com).

Electronical databases were searched from inception to December 15, 2020 for the combinations of the following terms: “light” [ALOIS], “light” AND “dementia” AND “actigraphy” [PUBMED], (“light” OR “illumination”) AND (“dementia” OR “Alzheimer”) AND (“actigraphy” OR “actimetry”) AND (“sleep” OR “circadian” AND “rhythm”) [SCOPUS], and “light” AND “dementia” AND “actigraphy” [GOOGLE SCHOLAR]. Furthermore, reference lists of six reviews were screened for additional references (Canazei et al., 2019; Cibeira et al., 2020; Dimitriou et al., 2017; Forbes et al., 2014; Hjetland et al., 2020; Mitolo et al., 2018).

Two authors (M. Canazei and E. M. Weiss) independently screened all titles and abstracts and retrieved the full text of any publication considered possibly eligible. Both authors then reviewed these publications against the eligibility criteria. Two authors (M. Canazei and I. Papousek) extracted data on the characteristics of the selected studies, including participants (number of participants, study population, study location, the severity of dementia, mean age, and percentage of female study participants), interventions (treatment and control intervention, light intensity, light exposure duration, and timing of light exposure), and outcomes (parametric and nonparametric circadian activity parameters and sleep parameters). The third author (E. M. Weiss) checked the extracted data for any errors. Finally, two authors (M. Canazei and I. Papousek) extracted outcome data from included studies and one author (M. Canazei) transferred data into the software Review Manager 5.4 (Review Manager, Cochrane Collaboration, 2020). All three authors independently assessed risks of bias for each included study by using the following seven criteria described in the Cochrane Handbook for Systematic Reviews of Interventions: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other bias (Higgins et al., 2011). In cluster-randomized trials, recruitment bias, baseline imbalance, loss of clusters, incorrect analysis, and comparability with individually randomized trials as further domains of bias were added. For cross-over trials, we added carry-over effect, availability of two periods of data, incorrect analysis, and comparability of results with those from parallel-group designs as further risks of bias. Any disagreement in the evaluation of the risks of bias between the authors was resolved by discussion. Afterward, we classified the overall risk of bias of a study as low if all domains were at low risk of bias, as high if at least one domain was at high risk of bias, or as unclear if at least one domain was at unclear risk of bias and no domain was at high risk (Higgins et al., 2011).

Data Analysis

Due to the small number of studies, all meta-analyses were run on pooled data from trials using various light interventions. Meta-analysis results for actigraphy outcomes were expressed as mean differences (MDs) with 95% confidence intervals (CIs). For multiarm trials, we selected appropriate study arms for the meta-analysis. For cluster-randomized trials, we followed recommendations from Campbell et al. (2001) and calculated the design effect based on an assumed intracluster correlation coefficient of 0.10.

Because the inclusion criteria for study participants were not restricted to any type and specific stage of dementia, clinical heterogeneity was likely to exist. As a consequence, we synthesized estimates of MD using random-effects meta-analysis models. The implementation of a subgroup analysis was not planned and thus not described in the preregistered study protocol. In the course of the systematic literature search, a significant number of eligible studies (four out of 12 randomized controlled studies) could be found where only people with Alzheimer’s disease were included. Moreover, there is initial evidence that bright light exposure may affect individuals with varying types of dementia differently (Mishima et al., 1998). Thus, we further pooled data from studies including persons with Alzheimer’s dementia only and ran meta-analyses for this subgroup.

The meta-analysis was primarily done with the inverse variance analysis method. We used the I² statistic (Higgins et al., 2011) to measure heterogeneity across trials in each analysis and considered heterogeneity as substantial when I² values were larger than 50%.

Results

The search of the four electronic databases, reviews, and reference lists of included studies resulted in 9,055 records with 4,553 articles remaining after the removal of duplicates (Figure 1). Of these, 4,521 articles were excluded after screening titles and abstracts, because studies did not meet selection criteria. Of the 36 retrieved articles, 23 studies were excluded after full-text review, 13 trials were identified as eligible for the systematic review, and 11 trials were subjected to meta-analyses (Table 1; excluded studies are referenced and reasons for exclusion are given in Supplementary Material S3).

Table 1.

Study Characteristics

Study (country)	Participants	Intervention and control groups	Outcome measures for meta-analyses	Major findings
Ancoli-Israel et al., 2002 (the United States)	N = 24 patients with various forms of dementia Location: three wards in two nursing homes Mean MMSE = 12.8 (range: 0–30) 75% female Mean age = 85.7 years	Intervention: light therapy device, 2,500 lux, cool-white light Control: adapted light therapy device, 50 lux red light Treatment: for 10 days, 2 h (from 5:30 to 7:30 p.m.)	Sleep parameters: not recorded Circadian activity parameters: COSINOR (amplitude, acrophase)	Bright light delayed acrophase
Ancoli-Israel et al., 2003 (the United States)	N = 61 AD patients with agitated behavior Location: nursing home Mean MMSE = 5.7 (range: 0–22) 68% female Mean age = 82.3 years	Intervention: light therapy device, 2,500 lux, white light Control: adapted light therapy device, <300 lux red light Treatment: for 10 days, 2 h (from 9:30 to 11:30 a.m.)	Sleep parameters: not included Circadian activity parameters: not recorded
Bromundt et al., 2019 (Switzerland)	N = 20 patients with various forms of dementia Location: nursing home Mean MMSE = 14.4 (range: 0–28.5) 85% female Mean age = 85.6 years	Intervention: dawn–dusk simulator with LED lamps with neutral-white light, dimming from 0.35 to 130 lux within 30 min Control: no active light intervention Treatment: for 8 weeks during individual sleep offset and onset periods	Sleep parameters: TST, SE, WASO Circadian activity parameters: NPCRA (IS, IV)	No effect on sleep parameters; no effect on circadian parameters
Burns et al., 2009 (the United Kingdom)	N = 48 patients with various forms of dementia with sleep disturbances and agitated behavior Location: two nursing homes Mean MMSE = 5.9 (SD: 5.5) 67% female Mean age = 83.4 years	Intervention: light therapy device, 10,000 lux, full-spectrum white light Control: adapted light therapy device, 100 lux full-spectrum white light Treatment: for 14 days, 2 h (from 10:00 a.m. to 12:00 p.m.)	Sleep parameters: TST Circadian activity parameters: not included	No effect on TST
Dowling et al., 2005 (the United States)	N = 46 AD patients with rest–activity disturbances Location: two long-term care facilities Mean MMSE = 7.0 (range: 0–23) 80% female Mean age = 84.0 years	Intervention: daylight or light therapy device, >2,500 lux, white light Control: 150–200 lux standard room light Treatment: for 10 weeks, 1 h (from 9:30 to 10:30 a.m.)	Sleep parameters: TST, SE, WASO Circadian activity parameters: COSINOR (amplitude, acrophase)	No effect on sleep parameters; light intervention stabilized acrophase
Dowling et al., 2008 (the United States)	N = 35 AD patients with rest–activity rhythm disruptions (e.g., insomnia, frequent nighttime awakenings, wandering, sundowning) Location: two long-term care facilities Mean MMSE = 9.3 (SD: 7.9) 86 % female Mean age = 86.0 years	Intervention: daylight or light therapy device, >2,500 lux, white light Control: 150–200 lux standard room light Treatment: for 10 weeks, 1 h (from 9:30 to 10:30 a.m.)	Sleep parameters: TST, WASO Circadian activity parameters: COSINOR (amplitude, acrophase), NPCRA (IS, IV)	No effect on sleep parameters; no effect on circadian parameters
Figueiro et al., 2019 (the United States)	N = 46 patients with various forms of dementia with sleep disturbances Location: four assisted-living and four long-term care facilities Mean MMSE = 15.6 (range: 4–24) 65% female Mean age = 85.1 years	Intervention: floor luminaries (550–600 lux and 5,000–7,000 K), light boxes (350 lux and 6,000 K), and light tables (750 lux and 5,000 K) generating each a circadian stimulus of 0.4 Control: floor luminaries (110 lux and 2,000–2,700 K), light boxes (100 lux and 2,700 K), and light tables (200 lux and 2,700 K) generating each a circadian stimulus <0.1 Treatment: for 4 weeks daily from individual awaking (6:00–8:00 a.m.) till 6:00 p.m.	Sleep parameters: TST, SE, SOL Circadian activity parameters: NPCRA (IS, IV)	No effect on sleep parameters; light intervention increased IV
Fontana-Gasio et al., 2003 (Switzerland)	N = 13 patients with various forms of dementia with sleep disturbances Location: two nursing homes and one clinic Mean MMSE = 13.9 (range: 2–21) 69% female Mean age = 85.0 years	Intervention: dawn–dusk simulator with halogen lamp, dimming from 0.001 to 210 lux Control: dawn–dusk simulator, red light, dimming from 0.001 to <5 lux Treatment: for 3 weeks, 30–34 min dawn, 30–44 min dusk, during individual sleep offset and onset periods	Sleep parameters: TST, SE, SOL Circadian activity parameters: not included	No effect on sleep parameters
McCurry et al., 2011 (the United States)	N = 67 AD patients with sleep problems Location: private homes Mean MMSE = 18.3 (SD: 7.0) 53% female Mean age = 80.9 years	Intervention: light therapy device, 2,500 lux, full-spectrum white light Control: no treatment Treatment: for 2 months, 1 h (within 2 h period before habitual bedtime)	Sleep parameters: TST, SE, WASO Circadian activity parameters: not recorded	No effect on sleep parameters
Mishima et al., 1998 (Japan)	N = 22 patients with various forms of dementia with sleep disturbances and agitated behavior Location: specialized ward Mean MMSE = 8.5 (range: 3–17) 59% female Mean age = 79.6 years	Intervention: light therapy device, 10,000 lux, white light Control: adapted light therapy device, 300 lux, white light Treatment: for 2 weeks, 2 h (9:00–11:00 a.m.)	Sleep parameters: not included Circadian activity parameters: not recorded
Riemersma-van der Lek et al., 2008 (The Netherlands)	N = 94 patients with various forms of dementia and participants with no dementia Location: 12 assisted care facilities Mean MMSE = 14.4 (SD: 6.6) 90% female Mean age = 85.8 years	Intervention: ceiling-mounted luminaires in living area, 1,000 lux white light Control: ceiling-mounted luminaires in living area, 300 lux white light Treatment: for 6 weeks, 9 h (9:00 a.m.–6:00 p.m.)	Sleep parameters: TST, SE, SOL, WASO Circadian activity parameters: not recorded	Light exposure increased TST
Sloane et al., 2007 (the United States)	N = 66 patients with various forms of dementia Location: two wards in a psychiatric facility and one ward in a dementia care facility Mean MMSE = 32% mild to moderate dementia, 68% severe dementia 47% female Mean age = 75.2 years	Intervention: ceiling-mounted luminaires in living area, 2,568 lux (measured horizontally at 1.2 m height) white light Control: ceiling-mounted luminaires in living area, 603 lux (measured horizontally at 1.2 m height) white light Treatment: for 3 weeks, 13 h (7:00 a.m.–8:00 p.m.)	Sleep parameters: TST Circadian rhythm parameter: COSINOR (amplitude, acrophase) and NPCRA (IV, IS)	Light exposure increased TST; light exposure advanced acrophase
Sloane et al., 2015 (the United States)	N = 17 patients with various forms of dementia with sleep disturbances Location: private homes Mean MMSE = 12.7 (SD: 9.1) 65% female Mean age = 77.2 years	Intervention: table and floor lamps in living area, 300–400 lux blueish-white light + light therapy device during dining, 100 lux, blue light Control: table and floor lamps in living area, 300–400 lux reddish-white light + light therapy device during dining, 100 lux, red light Treatment: for 6 weeks, approx. 11 h (from awakening [7:00 a.m.] to 6:00 p.m.)	Sleep parameters: TST, SE, SOL Circadian activity parameters: NPCRA (IV, IS)	No effect on sleep parameters; no effect on circadian parameters

Note: AD = Alzheimer’s dementia; MMSE = Mini-Mental State Examination; TST = total sleep time; SE = sleep efficiency; SOL = sleep onset latency; WASO = wake duration after sleep onset; IS = interdaily stability; IV = intradaily variability; NPCRA = nonparametric circadian rhythm analysis method; COSINOR = cosinor analysis.

Figure 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses study flow diagram.

Characteristics of the Included Trials

Study characteristics are summarized in Table 1 and described in full detail in Supplementary Material S1. Sample sizes in included randomized controlled trials varied from 13 (Fontana-Gasio et al., 2003) to 94 (Riemersma-van der Lek et al., 2008).

Study participants

The mean age of study participants ranged from 75.2 to 86.0 years. On average, studies included 70% women (range: 47%–90%). In four studies (Ancoli-Israel et al., 2003; Dowling et al., 2005, 2008; McCurry et al., 2011), only patients with Alzheimer dementia were included while in nine studies at least some participants were diagnosed with other forms of dementia, such as vascular dementia (Supplementary Material S1). In eight studies (Burns et al., 2009; Dowling et al., 2005, 2008; Figueiro et al., 2019; Fontana-Gasio et al., 2003; McCurry et al., 2011; Mishima et al., 1998; Sloane et al., 2015), patients additionally had sleep disturbances. Furthermore, in three studies (Ancoli-Israel et al., 2003; Burns et al., 2009; Mishima et al., 1998), study participants showed agitated behavior at inclusion. The severity of dementia of study participants was primarily assessed with the Mini-Mental State Examination (MMSE; Folstein et al., 1975) by primary care physicians, clinical psychologists, or gerontologists. Only in one study (Mishima et al., 1998), diagnostics was supplemented with neuroimaging techniques. In general, participants’ dementia stage varied considerably between the studies (mean MMSE scores ranged from moderate to severe).

Exclusion criteria for study participation differed substantially between the studies as well and comprised further neurological (e.g., sleep apnea), psychiatric (e.g., psychosis), somatic (e.g., stroke, hypertension, diabetes), or eye diseases (e.g., age-related macular degeneration, diabetic retinopathy), see Supplementary Material S1. Furthermore, patients receiving photosensitizing medication (McCurry et al., 2011) or suffering from motor disturbances (Ancoli-Israel et al., 2002; Mishima et al., 1998) were excluded from the study participation. In five studies (Ancoli-Israel et al., 2003; McCurry et al., 2011; Riemersma-van der Lek et al., 2008; Sloane et al., 2007, 2015), the ocular status of study participants was additionally checked at inclusion.

Medication was either stabilized before the study started (Mishima et al., 1998) or kept constant during the study (Dowling et al., 2005; Fontana-Gasio et al., 2003; McCurry et al., 2011; Riemersma-van der Lek et al., 2008; Sloane et al., 2007, 2015). Some studies excluded patients who were taking sedatives, hypnotics, or antipsychotics (Dowling et al., 2005, 2008; Mishima et al., 1998). Only in one study (Ancoli-Israel et al., 2003), authors did not report if and how the medication was dealt with.

Patients in two studies (McCurry et al., 2011; Sloane et al., 2015) lived in private homes, whereas in the other 11 studies patients either lived in nursing homes (Ancoli-Israel et al., 2002, 2003; Bromundt et al., 2019; Burns et al., 2009; Fontana-Gasio et al., 2003) or special care facilities (Dowling et al., 2005, 2008; Figueiro et al., 2019; Mishima et al., 1998; Riemersma-van der Lek et al., 2008; Sloane et al., 2007). Five studies implemented a single-center (Ancoli-Israel et al., 2003; Bromundt et al., 2019; McCurry et al., 2011; Mishima et al., 1998; Sloane et al., 2015) and eight studies a multicenter study design (Ancoli-Israel et al., 2002; Burns et al., 2009; Dowling et al., 2005, 2008; Figueiro et al., 2019; Fontana-Gasio et al., 2003; Riemersma-van der Lek et al., 2008; Sloane et al., 2007). Taking these factors together, the physical and mental status and the living situation of study participants varied considerably.

Light intervention

In four studies (Ancoli-Israel et al., 2002, 2003; Burns et al., 2009; Mishima et al., 1998), the light intervention period lasted up to 14 days, whereas in the other nine studies, patients were exposed to light from 3 weeks to 2 months. Seven trials (Ancoli-Israel et al., 2002, 2003; Burns et al., 2009; Dowling et al., 2005, 2008; McCurry et al., 2011; Mishima et al., 1998) examined the effects of light therapy with varying light intensities (2,500–10,000 lux) and exposure durations (1–2 h). Light therapies were applied from 10 days to 2 months in the morning (Ancoli-Israel et al., 2003; Burns et al., 2009; Dowling et al., 2005, 2008; Mishima et al., 1998), and in two studies in the evening, 2–3 h before patients typically fell asleep (Ancoli-Israel et al., 2002; McCurry et al., 2011).

In four studies (Figueiro et al., 2019; Riemersma-van der Lek et al., 2008; Sloane et al., 2007, 2015), room light interventions in bedrooms, living rooms, and common areas were implemented (Table 1). In these studies, either additional luminaires were installed to expose participants with at least 1,000 lux at eye level (Figueiro et al., 2019; Riemersma-van der Lek et al., 2008; Sloane et al., 2007), or research prototypes or light sources with an increased portion of short-wavelength light were used (Figueiro et al., 2019; Sloane et al., 2015). Daily ambient light exposure periods ranged from 9 to 13 h, and intervention periods were at least 3 weeks.

Finally, wall-mounted bedroom luminaires (so-called dawn–dusk simulators), which automatically increased ambient light levels either during awakening or decreased light levels while the person with dementia fell asleep, were used in two studies (Bromundt et al., 2019; Fontana-Gasio et al., 2003). Dawn–dusk simulators provided light over periods of approximately 30 min during participants’ sleep–wake transition with 130–210 lux (measured at the pillow toward the luminaire).

In six of the 13 studies (Ancoli-Israel et al., 2002, 2003; Burns et al., 2009; McCurry et al., 2011; Mishima et al., 1998; Riemersma-van der Lek et al., 2008), study participants were monitored during the light intervention to keep eyes open. In two of the six light therapy studies (Burns et al., 2009; McCurry et al., 2011), negative effects of light treatment were recorded.

Actimetry

In the reviewed studies, seven types of wrist-worn actimeters from five manufacturers were used. The sensitivity and sampling frequency of the accelerometers in these devices and the mode of deriving activity data were not described in the publications. Six studies did not specify the placement of the actimeter (Ancoli-Israel et al., 2002, 2003; Bromundt et al., 2019; Burns et al., 2009; Riemersma-van der Lek et al., 2008; Sloane et al., 2015). In two studies, the devices were worn on the dominant wrist (Dowling et al., 2005, 2008) and in the remaining five studies on the nondominant wrist.

In four studies (Ancoli-Israel et al., 2002, 2003; Fontana-Gasio et al., 2003; McCurry et al., 2011), staff reported a regular removal of actimeters during showering, and in further three studies (Dowling et al., 2005, 2008; Sloane et al., 2007) a special wristband (i.e., nylon lock cable) was used to avoid removal. Furthermore, in five studies (Ancoli-Israel et al., 2002, 2003; Bromundt et al., 2019; Fontana-Gasio et al., 2003; McCurry et al., 2011), visual inspection of activity raw data was described (with regard to prolonged periods of daytime inactivity), to exclude these periods from the analysis. Three of these studies (Bromundt et al., 2019; Fontana-Gasio et al., 2003; McCurry et al., 2011) explicitly described quantitative procedures to identify and substitute missing or erroneous activity data.

Research recommends at least 7 days of activity recordings for the analysis of circadian activity parameters (Camargos et al., 2013; van Someren, 2007). The length of actimetric recordings in eight of nine studies fulfilled this recommendation (Bromundt et al., 2019; Burns et al., 2009; Dowling et al., 2005, 2008; Figueiro et al., 2019; Fontana-Gasio et al., 2003; Sloane et al., 2007, 2015).

All but one study (Sloane et al., 2007) implemented a baseline measurement period. Ten studies measured effects on activity data once during light treatment, whereas one study (Bromundt et al., 2019) quantified light effects over two 4 weeks periods. Besides, in two studies (Burns et al., 2009; McCurry et al., 2011), light effects were recorded not during the treatment period but in the first postlight treatment days.

Deriving sleep parameters from activity data, which was done in all studies, requires information about the rising and bedtimes of study participants. While in seven studies (Supplementary Material S1) individual rising and bedtimes were derived from sleep diaries that were kept by staff members or caregivers, in the remaining six studies sleeping periods were habitual sleeping times in care facilities or hospitals.

Risk of bias across studies

Among the 13 randomized controlled trials, the main issues (Higgins et al., 2011) were “blinding of participants and personnel (performance bias)” and “other risks of bias,” for example, the inclusion of older adults with heterogeneous dementia diagnoses and recruitment of participants from different institutions with different care and treatment routines. Furthermore, information about random sequence generation and allocation concealment was unclear for the majority of trials (nine trials and seven trials, respectively). Details of the risk of bias assessment for included studies are provided in Figure 2.

Figure 2.

Review authors’ judgments about each methodological quality item for each included study. Note: The following abbreviations are used: “+” indicates a low risk of bias, “−” indicates a high risk of bias, and “?” indicates an unclear risk of bias.

The overall risk of bias is derived by adding the least favorable assessment across the domains of bias of every single study. In the present analysis, the overall risk of bias was high for 12 out of 13 studies and unclear for one study (Figueiro et al., 2019).

A summary of the proportion of trials that were at low, unclear, and high risk of bias for each domain is shown in Figure 3.

Figure 3.

Review authors’ judgments about each methodological quality item presented as percentages across all included studies.

Light intervention effects on sleep and circadian activity rhythm parameters

Parametric circadian activity rhythm parameters

Data from Ancoli-Israel et al. (2002), Dowling et al. (2005, 2008), and Sloane et al. (2007) were pooled to run meta-analyses on the two parametric circadian activity parameters amplitude and acrophase. We did not observe significant light intervention effects: amplitude (MD −0.02, 95% CI −0.08 to 0.05, I² = 0%, p = .62, n = 313) and acrophase (MD −0.24, 95% CI −0.72 to 0.25, I² = 0%, p = .34, n = 313). Forest plots of comparison are shown in Supplementary Figure S2.1 (amplitude) and Supplementary Figure S2.2 (acrophase).

Nonparametric circadian activity rhythm parameters

Data from five studies (Bromundt et al., 2019; Dowling et al., 2008; Figueiro et al., 2019; Sloane et al., 2007, 2015) were pooled for meta-analyses on two nonparametric circadian rhythm parameters. We did not find significant light effects on IV (MD 0.02, 95% CI −0.05 to 0.09, I² = 0%, p = .51, n = 354) and IS (MD −0.01, 95% CI −0.04 to 0.02, I² = 0%, p = .38, n = 354). Forest plots of comparison are shown in Supplementary Figure S2.3 (IV) and Supplementary Figure S2.4 (IS).

Sleep parameters

A meta-analysis on total sleep time summarized data from 10 studies (Bromundt et al., 2019; Burns et al., 2009; Dowling et al., 2005, 2008; Figueiro et al., 2019; Fontana-Gasio et al., 2003; McCurry et al., 2011; Riemersma-van der Lek, 2008; Sloane et al., 2007, 2015). We did not reveal a significant light effect on total sleep time in persons with dementia (MD 6.13, 95% CI −12.81 to 25.06, I² = 23%, p = .53, n = 594).

Pooled data on sleep efficiency from seven studies (Bromundt et al., 2019; Dowling et al., 2005; Figueiro et al., 2019; Fontana-Gasio et al., 2003; McCurry et al., 2011; Riemersma-van der Lek, 2008; Sloane et al., 2015) also did not yield significant light effects (MD 1.03, 95% CI −3.20 to 5.25, I² = 52%; p = .63, n = 333). Furthermore, duration of wake periods after sleep onset, pooled from four studies (Bromundt et al., 2019; Dowling et al., 2005, 2008; McCurry et al., 2011), was not affected by light interventions (MD 0.08, 95% CI −2.56 to 2.72, I² = 0%; p = .95, n = 212). Finally, pooling data from four studies on sleep onset latency (Figueiro et al., 2019; Fontana-Gasio et al., 2003; Riemersma-van der Lek, 2008; Sloane et al., 2015) also did not show a significant light effect (MD −4.66, 95% CI −12.80 to 3.49, I² = 5%; p = .26, n = 156). Forrest plots of comparison are shown in Supplementary Figure S2.5 (total sleep time), Supplementary Figure S2.6 (sleep efficiency), Supplementary Figure S2.7 (wake periods after sleep onset), and Supplementary Figure S2.8 (sleep onset latency).

Subgroup analyses including studies with persons with Alzheimer’s dementia

By pooling data from three studies including only persons with Alzheimer’s dementia (Dowling et al., 2005, 2008; McCurry et al., 2011), subgroup meta-analyses were run. No significant light effect was revealed for circadian activity parameters: amplitude (MD −0.05, 95% CI −0.36 to 0.25, I² = 0%; p = .73, n = 105), acrophase (MD −0.82, 95% CI −2.21 to 0.58, I² = 0%; p = .25, n = 105), IV (MD −0.16, 95% CI −0.42 to 0.10, p = .23, n = 35), and IS (MD −0.03, 95% CI −0.15 to 0.09, p = .63, n = 35). Furthermore, no light effects were found for sleep parameters: total sleep time (MD 4.30, 95% CI −32.03 to 40.63, I² = 0%; p = .82, n = 172), sleep efficiency (MD 0.93, 95% CI −4.46 to 6.32, I² = 0%; p = .73, n = 137), and duration of wake after sleep onset (MD 0.04, 95% CI −2.60 to 2.68, I² = 0%; p = .98, n = 172).

Discussion

This review examined adjunctive light intervention effects in persons with dementia using wrist actimetry. For the first time, circadian activity parameters were subjected to meta-analysis. Furthermore, we aimed at updating a Cochrane review from Forbes et al. (2014) on sleep-related outcome measures in persons with dementia. Including 11 trials, we found no evidence to support the notion that light stabilizes circadian activity rhythm parameters and improves nighttime sleep parameters. Subgroup analyses, including studies with persons with Alzheimer’s dementia, also did not demonstrate light effects on these outcome measures. The results are in line with findings from Forbes et al. (2014) who subjected seven studies to meta-analyses on four sleep parameters (total sleep time, sleep efficiency, sleep onset latency, and the number of nighttime awakenings) and showed no light effects. Six of the seven studies were included in the present review (Burns et al., 2009; Dowling et al., 2005, 2008; Fontana-Gasio et al., 2003; McCurry et al., 2011; Riemersma-van der Lek et al., 2008) and one was excluded because it was not published in a peer-reviewed journal (Nowak, 2008).

However, present results are in contrast to qualitative analyses from three recently published systematic reviews (Cibeira et al., 2020; Hjetland et al., 2020; Mitolo et al., 2018) including studies that were not restricted to randomized controlled study designs and actigraphy and showing some trends of beneficial light effects in persons with dementia. Mitolo et al. (2018) reviewed light intervention studies in patients with Alzheimer’s disease and reported mixed effects on subjective and objective parameters of sleep. Cibeira et al. (2020) published a systematic review on light therapy effects in older adults with cognitive impairment and reported some evidence for light effects on sleep parameters. Finally, Hjetland et al. (2020) summarized light interventions data in sleep and circadian activity parameters. Nine out of 15 included studies found improvements in some sleep outcomes, one study showed mixed results, and five found no effects. Furthermore, nine out of 13 studies reported significant circadian activity effects, while three found no effects.

We consider the quality of research of the included studies in this review as low for three reasons. All studies were at high or unclear risk of bias in at least one risk of bias domain and were heterogeneous in terms of light interventions and types and stages of dementia in included participants. Thus, the findings of this review should be interpreted with caution. In the following, we critically discuss two core study limitations with substantial practical implications: heterogeneity of light interventions and included study participants and study location.

Heterogeneity of Light Interventions

Seven out of the 13 included studies, published between 1998 and 2011 and thus the earlier ones, used a light therapy protocol. Interestingly, two studies followed common recommendations for light therapy (Ancoli-Israel et al., 2002, 2003; daily morning bright light exposure of 2,000 lux for 2 h), whereas in two studies light dose was higher (10,000 lux for 2 h, Burns et al., 2009; 6,200 lux for 2 h, Mishima et al., 1998) and in three studies lower (2,500 lux for 1 h, Dowling et al., 2005, 2008; McCurry et al., 2011) than recommended. The other six studies, published in the more recent period from 2007 to 2019, utilized ambient room lighting systems in dayrooms and bedrooms either for several hours during the day or during sleep–wake transition periods. Research indicates that compliance in light intervention studies with persons with dementia is low (Kim et al., 2003; van Someren et al., 1997), and attendance by staff or caregivers is necessary to ensure that persons with dementia stay in front of the light therapy device or in the brightly lit room and do not fall asleep during light exposure. It is noteworthy that in five of the seven light therapy studies, staff monitored the participants regularly during the interventions, whereas in the six studies using room lighting systems, participants were monitored only in one study (Riemersma-van der Lek et al., 2008). Longer exposure durations under room light basically allow people to behave more freely and naturally during light exposure, and attendance is less necessary.

It is well documented (Brouwer et al., 2017; Kogan & Guilford, 1998; Terman & Terman, 1999) that up to 45% of younger persons using light therapy devices for the treatment of seasonal affective disorders reported transient ocular discomfort and vision problems. So far, little attention has been paid to the negative side effects of light therapy in studies with persons with dementia. Only two of six included light therapy studies (Sloane et al., 2007, 2015) recorded adverse effects.

Furthermore, light therapy is contraindicated when patients are taking photo-sensitizing medication or suffer from visual impairments (e.g., cataract, macular degeneration). Because this occurs frequently in persons with dementia, this exclusion criterion significantly reduces possible recipients of light therapy interventions. Taken together, the necessity of increased staff or caregiver attendance and a potentially adverse side-effect profile hampers a broad application of light therapy in persons with dementia. In contrast, room light interventions (e.g., increased ambient light levels in dayrooms and dawn/dusk simulators in bedrooms) expose persons to lower light intensities and, thus, decrease the risk of adverse light effects.

Light intensity, timing, and exposure duration of light interventions in included studies varied considerably. Included light therapy studies that had exposed patients to at least 2,500 lux at eye level, had short exposure durations of up to 2 h. By contrast, room light interventions in dayrooms with light intensities of maximally 1,000 lux at eye level lasted from 9 to 13 h. More research is needed to elucidate the relationships between light intensity, timing, and exposure duration to support efficient light treatment in persons with dementia (Kim et al., 2003; Praschak-Rieder & Willeit, 2003).

Finally, controlling the amount of light persons with dementia are exposed to during light treatment is difficult but essential to ensure sufficient photic stimulation. Small-surface light therapy devices are particularly pertinent to this topic (Oldham et al., 2019). In general, the intensity of photic stimulation decreases with the square of the distance between the person and the light therapy device. As a consequence, illuminance levels of light therapy devices are often insufficient when the distance exceeds 1 m (Oldham et al., 2019). Controlling ambient light exposure levels in living rooms and dayrooms over an extended period is difficult as well, and practically impossible when using a dawn–dusk simulator in bedrooms while persons are in a sleep–wake transition state. Controlling the intensity of the light stimulus is thus one of the most difficult tasks in light intervention research, and it was only implemented in two of the included room light intervention studies (Figueiro et al., 2019; Sloane et al., 2007).

Heterogeneity of Study Participants and Homogeneity of Study Locations

Four studies of this review (Ancoli-Israel et al., 2003; Dowling et al., 2005, 2008; McCurry et al., 2011) included patients with Alzheimer’s dementia. In the remaining nine studies, participants with various types of dementia were enrolled. Also, the severity of dementia varied considerably, and in 11 studies participants were included because sleep problems and agitated behaviors co-occurred. There is the first evidence that light intervention may affect individuals with varying health conditions and types of dementia differently. For instance, in one included study, light therapy reduced nighttime activity in patients with vascular dementia but did not affect patients with Alzheimer’s dementia (Mishima et al., 1998). Furthermore, the response to light interventions might be moderated by dementia severity, but the direction of response is not clear yet (Hjetland et al., 2020).

Studies including older adults with moderate to severe dementia who live in care facilities face the challenge that their daily activities are structured to a significant extent and sleeping, eating, socializing, care, and medical treatments occur at specific times of the day. These routines might significantly influence the timing and magnitude of daytime physical activities (Kredlow et al., 2015) and the timing and duration of nighttime sleep periods (Reid & Zee, 2005). In 11 of the 13 included studies, persons with dementia lived in nursing homes, care facilities, or psychiatric wards. None of these studies described the daily routines of the participants. In five studies (Ancoli-Israel et al., 2002; Dowling et al., 2005, 2008; Mishima et al., 1998; Sloane et al., 2007), fixed institutional bedtimes were used for sleep analyses, which indicates that activities of daily living were not entirely self-determined by participants in care facilities. In addition to light, social rhythms might have influenced the timing of the circadian system and acted as strong Zeitgeber.

Limitations

This review has four main limitations. First, 36 studies were found eligible for inclusion but only 13 of these studies were incorporated for analyses. One prominent reason for exclusion was that studies had not implemented a randomized, controlled trial design. By rigorously selecting high-quality research, we limited the database for our analyses and did not find evidence for beneficial light effects. Second, sample sizes and the number of trials that examined each outcome parameter were small. Results thus should be interpreted with caution. Third, both, study populations and light interventions varied considerably. As a consequence, we did not run meta-analyses on the three different types of light interventions (i.e., light therapy, room lighting during the day, and dawn–dusk simulation). Furthermore, only one subgroup analysis was carried out pooling data from three studies including persons with Alzheimer’s dementia. Fourth, there are a few limitations in subgroup analysis. This analysis was not included in the registered study protocol. Moreover, it was carried out according to diagnosis, although some studies did not specify the exact type of dementia diagnosis. In addition, subgroup analysis was not carried out according to risk of bias.

Conclusions

Bright light exposure has a wide range of applications in the field of mental health. There is good evidence that light therapy is effective in seasonal and nonseasonal affective disorders (Pjrek et al., 2020; Tao et al., 2020). Bright light exposure may further alleviate symptoms of sleep disorders (van Maanen et al., 2016), circadian rhythm sleep disorders (Faulkner et al., 2019), or eating disorders (Beauchamp & Lundgren, 2016). Even there is some evidence for beneficial light effects in the healthy population, for instance, in reducing symptoms of jetlag (Bin et al., 2019) and improving adjustment to shifting working routines (Aemmi et al., 2020). Besides, in older adults aged 60 and older, light exposure may improve mood in nonseasonal depression (Zhao et al., 2018) and sleep quality (Montgomery & Dennis, 2004). Interestingly, the mechanisms of action of light treatment remain unclear. Studies indicate that bright light affects the monoaminergic system (e.g., serotonin, norepinephrine, and dopamine), the circadian system (e.g., by shifting the phase or altering the amplitude of circadian parameters), and the autonomous nervous system (Oldham & Ciraulo, 2014) and thus may improve sleep and mood and stabilize the circadian system (Musiek et al., 2015).

Based on this review, there is still insufficient evidence to justify the use of adjunctive light interventions for the improvement of sleep in persons with dementia. Additionally, we found no evidence for light effects on a stabilized circadian activity rhythm. Further high-quality research with larger sample sizes and more carefully selected study populations is needed. Light impact research in this domain is highly multidisciplinary. Future research teams should comprise experts in the fields of geriatric care and medicine, lighting design and engineering, and wearable sensor technology.

Practical Recommendations

There is no convincing evidence for the beneficial effects of adjunctive light on circadian activity and sleep parameters in persons with dementia. More research is necessary before specific evidence-based recommendations on light interventions can be given. High-quality research may further help to specify the optimal timing, duration, light intensity, and color spectrum of bright light treatment and may help to establish bright light exposure as an effective nonpharmacological additional treatment option for persons with dementia and may even influence future lighting design standards in geriatric clinics and care facilities.

Funding

This work was supported by the Austrian Research Promotion Agency (grant number: 850747).

Conflict of Interest

None declared.