Association of Sleep with Cognition and β-Amyloid Accumulation in Adults with Down Syndrome

Abstract

Adults with Down syndrome have an increased risk for both disordered sleep and Alzheimer’s disease (AD). In the general population, disrupted sleep has been linked to β-amyloid accumulation, an early pathophysiologic feature of AD. In this study, the association between sleep, β-amyloid, and measures of AD-related cognitive decline were examined in 47 non-demented adults with Down syndrome (aged 26–56 years). Sleep was measured using actigraphy over 7 nights. Pittsburgh Compound B (PiB) PET was used to assess global and striatal β-amyloid burden. Participants had the following clinical AD status: 7(15%) mild cognitive impairment and 40 (85%) cognitively unaffected. Average length of nighttime awakenings was significantly positively associated with striatal β-amyloid and decreased cognitive performance in executive functioning and motor planning and coordination. Findings suggest that disrupted sleep is associated with β-amyloid accumulation and cognitive features of preclinical AD in Down syndrome. Early identification and treatment of sleep problems could be a lifestyle intervention that may delay β-amyloid accumulation and cognitive decline in this AD vulnerable group.

1. Introduction

Sleep may play a role in Alzheimer’s disease (AD). Recently, sleep disruption and low sleep efficiency have been associated with β-amyloid accumulation (Bubu et al., 2017; Ju et al., 2013), an early pathophysiologic hallmark of AD (Jack et al., 2013). In cognitively healthy older adults, β-amyloid accumulation has been associated with self-reports (Carvalho et al., 2018; Spira et al., 2013; Sprecher et al., 2015; Sprecher et al., 2017) of poor sleep quality, sleep problems, and daytime drowsiness, as well as actigraph (Ju et al., 2013) measured levels of sleep efficiency. These findings suggest that sleep disruptions may be implicated in the preclinical stages of AD (Sperling et al., 2009), when β-amyloid accumulation occurs prior to the onset of AD dementia.

One population (Mann & Esiri, 1989; Wisniewski et al., 1978; Zigman & Lott IT, 2007) with an increased incidence and early onset of AD is adults with Down syndrome (DS). Occurring in 1 in 691 live births in the U.S. (Parker et al., 2010), DS is a developmental disability most commonly due to full triplication of chromosome 21. The triplication of the amyloid precursor protein (APP), located on chromosome 21, is purported (Wiseman et al., 2015) to result in the over production of β-amyloid and lead to the increased risk for AD. Nearly all adults with DS exhibit β-amyloid plaque burden by 40 years of age (Wiseman et al., 2015), and more than three-fourths of adults with DS exhibit clinical AD dementia by age 65 years (McCarron et al., 2017). Despite the shared genetic risk for AD, there is variability in the age of onset of β-amyloid accumulation and in the age of onset of clinical AD dementia in DS (Handen et al., 2012; Lao et al., 2016; Lao et al., 2017). Sleep could be a modifiable lifestyle factor related to this variability and of particular relevance to DS. Approximately, one-half of adults with DS exhibit sleep disordered breathing, especially obstructive sleep apnea (Trois et al., 2009) and one-third (Stores, 2019) experience difficulties initiating sleep, nighttime awakenings, talking in sleep, and/or restless sleep. To date, no published research has examined whether sleep disruptions are associated with β-amyloid accumulation in DS or are related to early and subtle cognitive declines in the preclinical and mild cognitive impairment (MCI) stages of AD in DS.

The objective of the current study was to examine the association between sleep, β-amyloid using [C-11] Pittsburgh Compound-B (PiB) PET imaging, and AD-related domains of cognitive functioning in 47 non-demented adults with DS. Sleep was assessed across seven nights via a wrist-worn actigraph accelerometer, as well as a sleep record jointly reported on by the adult with DS and a caregiver. PiB PET β-amyloid was assessed globally, as well as specifically in the striatum due to the frequent, early, and predominant pattern of striatal β-amyloid accumulation in DS (Handen et al., 2012; Lao et al., 2016; Lao et al., 2017). We examined AD-related domains of cognitive functioning commonly affected (Ball et al., 2008; Hartley et al., 2017; Krinsky-McHale et al., 2008; Startin et al., 2019) in the preclinical and MCI stages of AD in DS: memory, executive functioning, and motor planning and coordination. We hypothesized that disrupted sleep in adults with DS would be associated with a higher level of β-amyloid and worse cognitive performance.

2. Methods

2.1. Participants

Forty-seven non-demented adults with DS (aged 26 to 56 yrs) with confirmed full trisomy 21 participated in the study. Participants were recruited from the Alzheimer’s Biomarker Consortium in DS (ABC-DS; https://www.nia.nih.gov/research/abc-ds), a longitudinal multi-site study of AD in DS. As part of the ABC-DS study, adults with DS undergo comprehensive cognitive testing, caregiver interviews, medical history reviews, and MRI and PET imaging. At the University of Wisconsin-Madison, ABC-DS participants were invited to join an auxiliary study that included assessment of sleep. Study inclusion criteria for the auxiliary sleep study included: age ≥ 25 years, mental age of ≥ 30 months, genetic testing confirming trisomy 21, no conditions contraindicative for neuroimaging scans (e.g., pregnant or breastfeeding, metal in the body, claustrophobia or behavioral concerns), no medical or psychiatric conditions that impair cognitive functioning, and no clinical diagnosis of AD dementia. In total, 47 of the 50 adults with DS who met inclusion criteria for the auxiliary sleep study, enrolled in the study. The 7-night assessment of sleep occurred concurrently with (n = 22, 47%) or up to 1.5 years after (n = 25, 53%) the ABC-DS study visit (occurring over two days) involving the assessment of cognitive functioning, MRI, and PiB PET, with a mean lapse of 6.4 months (SD = 7.1). Participant characteristics are summarized in Table 1.

Table 1.

Participant Characteristics

	Total (N=47)	Cognitively unaffected (N=40)	MCI (N=7)	Group p-valuea	Effect sizeb
Female, No. (%)	24 (51.1)	22 (55.0)	2 (28.6)	0.197	-
Age at sleep assessment, y	38.4 (8.3)	36.4 (7.1)	50.0 (5.0)	0.001	2.215
Age at PiB PET scan, y	37.9 (8.2)	35.8 (6.8)	49.6 (4.7)	0.001	2.361
Interval sleep assessment to PiB PET, y	0.5 (0.6)	0.6 (0.6)	0.4 (0.5)	0.432	-
Mental age equivalentc, y	8.0 (3.4)	8.2 (3.5)	6.5 (2.0)	0.199	-
PiB Positive, No. (%)	9 (19.1)	4 (10.0)	5 (71.4)	0.001	0.556bb
PiB Striatum SUVR	1.3 (0.3)	1.2 (0.2)	1.6 (0.4)	0.014	1.265
PiB Global SUVR	1.1 (0.2)	1.1 (0.2)	1.4 (0.2)	0.012	1.500
APOE e4 positived, No. (%)	8 (17.8)	7 (18.4)	1 (14.3)	0.793	-
Body mass indexe	33.0 (8.1)	33.7 (8.1)	28.3 (7.2)	0.129	-
Caregiver report of sleep apnea, No. (%)	17 (36.2)	15 (37.5)	2 (28.6)	0.650	-
Use of any sleep medication, No. (%)	3 (6.4)	3 (7.5)	0	1.000	-
Actigraph sleep measures:
Total sleep time, hr	6.8 (1.1)	6.9 (1.2)	6.3 (0.9)	0.178	-
Wake after sleep onset, min	113.4 (41.2)	109.2 (40.0)	137.7 (42.2)	0.091	-
Sleep efficiencyf, %	75.4 (7.6)	76.3 (7.5)	70.0 (6.9)	0.043	0.874
Sleep fragmentation indexg, %	44.5 (11.1)	43.6 (10.5)	49.7 (13.7)	0.183	-
Number of awakenings, n	23.3 (8.1)	23.9 (8.3)	19.8 (5.6)	0.223	-
Length of awakenings, min	5.2 (2.9)	4.7 (2.8)	7.9 (2.2)	0.007	1.271

Note: Unless otherwise indicated, data are expressed as mean (SD).

APOE = apolipoprotein E; SUVR = standard uptake value ratio.

^aIndependent samples t-tests, χ², or Fisher’s exact test of group differences.

^bCohen’s for significant t-test group differences

^bbCramér’s for significant χ² group differences

^cMental age equivalent was derived from the Peabody Picture Vocabulary Test, Fourth Edition.

^dAPOE allele information was missing for 2 cognitively unaffected subjects.

^eCalculated as weight in kilograms divided by height in meters squared.

^fCalculated³⁷ as 100% × sleep duration divided by the total time spent in bed.

^gCalculated³⁷ as 100% × the number of groups of consecutive mobile 30 second epochs divided by the total number of immobile epochs.

2.2. Measures

2.2.1. Overall cognitive ability

Participants’ overall cognitive ability was assessed using the Peabody Picture Vocabulary Test-Fourth Edition (PPVT-4; Dunn & Dunn, 2007), a measure of receptive language ability that is reliable and valid in adults with intellectual disability and has been highly correlated with overall intellectual ability in DS (Phillips et al., 2014). The PPVT-4 mental age equivalent score was used to control for differences in overall cognitive ability. The remaining cognitive assessments were administered in the same order for every participant (i.e., no counterbalancing).

2.2.2. Clinical AD Status

Determination of clinical AD status was based on a case consensus process involving three to four research staff (including a licensed psychologist and physician) experienced in AD in DS and blind to neuroimaging and sleep data. The following information from all available study visits (participants had 2 to 5 visits) in the ABC-DS study was used in the case consensus process and considered in reference to overall cognitive ability and psychiatric and medical conditions: a) Down Syndrome Mental Status Examination (DSMSE; Haxby, 1989), a directly-administered measure of dementia symptoms including recall of personal information, orientation to season and day of week, immediate and delayed memory, language, visuospatial function and praxia; b) Dementia Questionnaire for People with Learning Disabilities (DLD; Evenhuis, 2018), an informant questionnaire assessing short-term memory, long-term memory spatial and temporal orientation, speech, practical skills, mood, activity and interest, and behavioral disturbances; c) Vineland Adaptive Behavior Scales, Third Edition (Vineland-3; Sparrow et al., 2016) an informant report of daily living skills; d) performance on three directly-administered measures – Developmental Test of Visual-Motor Integration, 5th Edition (Beery, Buktenica, & Beery, 2004), Wechsler Intelligence Scale for Children [Wechsler, 2004] and Block Design and Haxby extension (Haxby, 1989), and Developmental NEuroPSYchological Assessment (Korkman, 1998) Word Generation Semantic Fluency subtest; and; e) caregiver report of recent life events (e.g., staff or residential changes). Clinical AD status categories were: a) cognitively unaffected; b) MCI-DS, indicating mild cognitive declines limited in scope; c) dementia, indicating marked cognitive decline and decreases in daily functioning; and d) unable to determine.

2.2.3. Working and Episodic Memory

The Cued Recall Test measures learning and episodic verbal memory and has been demonstrated (Zimmerli & Devenny, 1995) to be reliable and valid in adults with DS. In the learning phase, individuals attempt to learn 12 words (e.g., names of objects in pictures) that are linked to categories. The Free and Cued Recall score is the sum of the number of pictures accurately recalled in three free recall trials and three cued recall trials (i.e., category given). The Cued Recall Intrusions score is the number of incorrectly recalled pictures in the cued recall trials.

The Picture Recognition subtest of the Rivermead Behavioral Memory Test (RBMT; Wilson et al., 1991) measures episodic visual memory and has been negatively associated with β-amyloid accumulation in DS (Hartley et al., 2014; Hartley et al., 2017). For the RBMT, individuals view a series of ten pictures, and then after a delay, they are shown twenty pictures (the ten previous plus ten new pictures) and asked to recall which pictures they have seen previously. Current analyses used the total number of correctly remembered pictures minus the number of false positives (e.g., recalling a picture that was not previously shown).

The Wechsler Memory Scale, 4th Edition (WISC-IV; Wechsler, 2004) Digit Span Forwards, a task previously shown to have strong construct validity in adults with DS (Devenny et al., 2005), requires individuals to correctly recall a string of digits in the correct order (forward). There are a total of 8 possible items, each with two trials. The first item is composed of a 2-digit string, and items become increasingly difficult (i.e. digit-string length increases). Each trial is scored as either 0 (did not recall correctly) or 1 (recalled correctly). The assessment is discontinued after the participant scores a 0 for both trials of an item. The raw score used in the current analysis was the number of correctly recalled trials.

2.2.4. Executive Functioning

The WISC-IV Digit Span Backwards requires individuals to correctly remember, and then reverse the order of, a string of digits. Identical to the scoring for the WISC-IV Digit Span Forwards (see above), there are 8 potential items, each with 2 trials. The raw score, determined by summing the number of correctly trials, was used in current analyses. The assessment is discontinued after the participant scores a 0 for both trials of an item.

The Cat and Dog Modified Stroop Task (Nash & Snowling, 2008) assesses executive functioning. Participants view a series of 16 pictures of cats and dogs and are asked to name the pictures as quickly as possible (Naming trial). Participants are then instructed to switch the names of the pictures (e.g., say “cat” when referring to a picture of a dog; Switch trial). The Cat-Dog Switch Time is the time in seconds taken during the Switch trial and this score was correlated with the number of errors in the Switch trial (r =.562 p = .001) in the current sample. These measures have been established (Ball et al., 2008; Devenny et al., 2005) as reliable and valid in previous studies of adults with DS.

2.2.5. Motor Planning and Coordination

The Purdue Pegboard (Vega, 1969) assesses fine motor planning and coordination by having individuals place pegs into holes on a pegboard. There are three 30 second trials that are administered in the following order: dominant hand, non-dominant hand, and both hands simultaneously, with each trial using the number of pegs placed in holes as the measured outcome. The single-hand scores have been negatively associated with β-amyloid accumulation in non-demented adults with DS (Hartley et al., 2017), and the both hands score has been found to be associated with measures of motor coordination and executive functioning in adults with intellectual disability (Burt et al., 2005). In current analyses, we focused on the both hands score.

2.3. MRI Acquisition

Participants were scanned on a GE 3T MR750 using an 8-channel head coil. A T1-weighted brain image was acquired in the axial plane with a high resolution volumetric-spoiled gradient sequence (inversion time/echo time/repetition time = 450/3.2/8.2 ms, flip angle = 12°, slice thickness = 1 mm no gap, and matrix size = 256 × 256 × 156). MRI data were used for PET-MRI registration and brain region definition.

2.3.1. PiB PET Acquisition

[C-11] PiB PET radiochemical synthesis, acquisition parameters, and generation of standard uptake value ratios (SUVRs) were detailed previously (Cohen et al., 2013). Briefly, the PiB PET scan was acquired for 40–70 minutes post-injection. The reconstructed images were then corrected for inter-frame motion. Parametric SUVR images were generated from 50 to 70 minute PET frames using cerebellar gray matter as a reference region. PET images were spatially normalized to MNI152 space using a DS-specific PiB PET template (Lao et al., 2019).

2.3.2. Image Processing

Regions of interest were defined in MNI152 space using the Wake Forest University PickAtlas toolbox in SPM. The investigated regions of interest included the frontal cortex, anterior cingulate gyrus, parietal cortex, temporal cortex, precuneus, and striatum. An average global standard update value ratio (SUVR) was computed across the regions of interest (ROIs). Cut-offs for ROI-specific PiB positivity were determined by sparse k-means clustering with resampling using a previously described process (Cohen et al., 2013; Lao et al., 2019). Participants were classified as PiB positive (PiB(+)) if at least one ROI exceeded its threshold (anterior cingulate, 1.59; frontal cortex, 1.48; parietal cortex, 1.51; precuneus, 1.64; striatum, 1.45; and temporal cortex, 1.38).

2.4. Sleep Assessment

Participants were instructed to wear an ActiGraph GT9X Accelerometer on their non-dominant wrist continuously over a 7-night period. Acitgraphs were programmed to sample at a rate of 30Hz and the epoch length was 60 seconds. The ActiLife 6 software (Version 6.13.1, ActiGraph) was used to analyze data and sleep-wake identification was determined using the Cole-Kripke algorithm (Cole & Kripke, 1992), which involves a weighted determination of sleep vs. wake for each epoch relative to the four previous and two future epochs. In adult populations, the GTX series wrist-worn actigraphs using the Cole-Kripke algorithm have been shown to be comparable (Quante et al., 2018) to other leading wrist-worn devices and had adequate concurrent validity with data from polysomnography. Actigraph variables of interest included: a) total sleep time (TST) in minutes; b) wake after sleep onset (WASO), defined as minutes awake after initially falling asleep; c) sleep efficiency (SE), defined as the minutes asleep divided by the total number of minutes in bed; d) sleep fragmentation index (SFI), defined as the proportion of total sleep epochs characterized by movement; e) number of awakenings (NOA), defined as number of times woke up; f) length of awakenings (LOA), defined as the average length of awakenings in minutes. The mean level (i.e., average across the 7 nights) was used in analyses for all actigraph indices.

Participants and a caregiver also jointly completed a daily sleep record for the 7 nights. This data was used to analyze and validate the actigraph data. For all 7 days/nights, in the daily sleep record, participants and caregivers jointly reported the time the participant got in and out of bed, number of times the participant woke up during the night and number of minutes the participant was awake after falling asleep, and number of daytime naps.

2.5. Statistical Analysis

Descriptive statistics were first used to examine clinical AD status, PiB PET status (i.e., +/−), and mean and range for actigraph accelerometer sleep variables of study participants. Pearson correlations were used to examine the association between actigraph sleep indices and the daily sleep records reported on by the adult with DS and their caregiver. The association between global and striatal PiB SUVRs and actigraph sleep indices were examined using Pearson correlations. Associations between actigraph sleep indices and memory, executive functioning, and motor planning and coordination were examined using partial correlations that controlled for overall cognitive ability (PPVT-4). Given the strong association between chronological age, cognitive functioning, and β-amyloid accumulation in DS (Handen et al., 2012; Hartley et al., 2017; Lao et al., 2016) the above analyses were re-run adjusting for chronological age. Finally, to understand if associations between sleep, β-amyloid, and cognitive functioning begin prior to MCI-DS, we also re-ran analyses with the cognitively unaffected participants (N=40) and ran independent sample t-tests to compare actigraph sleep indices between the cognitively unaffected and MCI-DS participants.

3. Results

3.1. Descriptive Information and Preliminary Analyses

In this cohort of adults with DS (Table 1), the average chronological age at the time of the sleep assessment was 38.4 years (SD = 8.3) and the average receptive language mental age equivalent was 8.0 years (SD = 3.4). Approximately half of the participants were female (n = 24, 51%). Twenty participants had a sleep disorder according to caregiver report; 17 (36%) participants had sleep apnea (of whom 10 were being actively treated), and 3 participants had difficulties falling and/or remaining asleep (all of whom were taking at least one sleep medication). Reported sleep medications included Tylenol PM, Provigil, Lorazepam, Trazodone, and Prazosin. Nine (18%) participants were APOE e4 carriers. Seven (15%) participants received a clinical status of MCI-DS and 40 (85%) participants received a clinical status of cognitively unaffected. No participants were diagnosed with clinical dementia. Nine participants (19%) were PiB(+), including all MCI-DS participants and 2 cognitively unaffected participants. On average, participants had less than 7 hours of sleep per night and woke up 23 times for an average length of 5 minutes. Forty-three (91%) participants had actigraph data for ≥ 6 nights. Forty-five (96%) participants (and their caregivers) completed the sleep record for all 7 nights, and two participants (and their caregivers) completed it for 6 nights. Of the remaining 4 participants, data was included if they reported a minimum of 4 nights. The mean TST (i.e., average across the 7 nights) on the actigraph was significantly positively associated with the mean time spent in bed (r = .666, p <.001) in the sleep record. The mean WASO on the actigraph was significantly positively associated with mean number of minutes awake reported in the sleep record (r = .319, p < .05). There were non-significant correlations between the other actigraphy and related indices reported in the sleep diary (r(s) = .016 to .289, p >.05), which is in line with previous reports that objective measures (e.g., actigraph and polysomography) capture sleep disruptions that are often not evident in sleep diaries (Gimenez et al., 2018).

3.2. PET β-Amyloid and Sleep

In Pearson correlations between PiB PET global and striatal SUVRs and actigraph sleep indices, a higher level of striatal PiB was modestly associated with LOA (r =.323, p < .027) (Table 2). There were no additional significant associations. When adjusting for chronological age, there were no significant associations between global or striatal PiB and actigraph sleep indices (Table 2). Similarly, when examining PiB and actigraph measures in the cognitively unaffected subset (N=40), the association between striatal PiB and mean LOA was not significant (r =.269, p = .093; Table 3).

Table 2.

Association between striatal and global PiB PET β-Amyloid and actigraph sleep indices in the full cohort of non-demented individuals with DS (N=47)

			Pearson correlation (r)						Adjusting for chronological age Partial correlation (r)
			TST	WASO	SE	SFI	NOA	LOA	TST	WASO	SE	SFI	NOA	LOA
PiBPET Amyloid	Striatum SUVR	Corr.	−.022	.012	−.030	−.007	−.115	.323	.006	−.100	.109	−.076	.064	−.010
		Sig.	.882	.938	.840	.960	.441	.027	.970	.507	.472	.615	.671	.949
	Global SUVR	Corr.	.043	.086	−.007	−.027	−.116	.262	.080	.031	.093	−.091	.001	.000
		Sig.	.772	.564	.965	.855	.436	.075	.596	.836	.538	.545	.946	.998

Note. TST = total sleep time; SE = sleep efficiency; WASO = wake after sleep onset; SE = sleep efficiency; SFI = sleep fragmentation index; NOA = number of awakenings; LOA = length of awakenings; SUVR = standard uptake value ratio; PiB = Pittsburgh Compound B; Sig. = Significance (2-tailed); Corr. = Correlation. Bolded text, p <.05

Table 3.

Association between striatal and global PiB PET β-Amyloid and actigraph sleep indices in the cognitively unaffected subset (N=40)

			Pearson correlation (r)						Adjusting for chronological age Partial correlation (r)
			TST	WASO	SE	SFI	NOA	LOA	TST	WASO	SE	SFI	NOA	LOA
PET PiBAmyloid	Striatum SUVR	Corr.	.082	−.136	.114	−.035	−.090	.269	.005	−.126	.093	.027	.020	.081
		Sig.	.616	.402	.484	.831	.581	.093	.974	.443	.574	.830	.905	.624
	Global SUVR	Corr.	.155	−.070	.165	−.093	−.107	.163	.118	−.047	.150	−.065	−.045	.030
		Sig.	.340	.667	.310	.569	.513	.316	.475	.778	.361	.695	.787	.854

Note. TST = total sleep time; SE = sleep efficiency; WASO = wake after sleep onset; SE = sleep efficiency; SFI = sleep fragmentation index; NOA = number of awakenings; LOA = length of awakenings; SUVR = standard uptake value ratio; PiB = Pittsburgh Compound B; Sig. = Significance (2-tailed); Corr. = Correlation.

3.3. Memory, Executive Functioning, and Motor Planning and Coordination and Sleep

Partial correlations, adjusting for overall cognitive ability (PPVT-4), were used to examine the association between memory, executive functioning, and motor planning and coordination and actigraph sleep indices (Table 4). Poorer performance across all three cognitive domains was related to greater mean LOA. Specifically, a lower Purdue Pegboard score (r = −.302, p = .044) and a higher Cat Dog Switch Time (r = .358, p = .016) were significantly associated with higher mean LOA. In addition, there was a weak positive correlation between Cued Recall Intrusion and mean LOA (r = .291, p = .052). There were no other significant associations. After adjusting for chronological age, only the positive association between LOA and the Cat Dog Switch Time remained significant (r = .299, p = .049; Table 4).

Table 4.

Association between AD-related cognitive domains and actigraph sleep indices in the full cohort of non-demented individuals with DS (N=47)

			Adjusting for Mental Age						Adjusting for Mental Age and Chronological Age
			TST	WASO	SE	SFI	NOA	LOA	TST	WASO	SE	SFI	NOA	LOA
Working Memory	Free and Cued	Corr.	.088	−.283	.256	−.123	.094	−.259	.081	−.265	.225	−.104	.180	−.132
		Sig.	.568	.059	.090	.420	.533	.086	.602	.082	.142	.503	.243	.393
	Cued Recall Intrusions	Corr.	−.041	.239	−.208	.121	.049	.291	−.032	.218	−.174	.102	.127	.175
		Sig.	.789	.114	.171	.428	.747	.052	.838	.155	.258	.510	.410	.257
	Rivermead	Corr.	−.072	.073	−.132	.065	.120	.156	−.075	.083	−.146	.073	.104	−.132
		Sig.	.639	.634	.386	.670	.434	.305	.628	.593	.344	.640	.502	.393
	Digit Span Forward	Corr.	.111	−.004	.074	−.087	−.171	−.066	.110	.001	.069	−.084	−.185	−.050
		Sig.	.466	.978	.627	.570	.262	.667	.447	.996	.658	.590	.228	.748
Executive Functioning	Digit Span Backward	Corr.	−.060	.048	.086	−.068	.286	−.154	.056	.063	.069	−.059	.267	−.108
		Sig.	.696	.755	.574	.656	.056	.311	.718	.687	.655	.705	.080	.486
	Cat and Dog	Corr.	−.179	.074	−.153	.110	.009	.358	−.176	.052	−.128	.096	.058	.299
		Sig.	.239	.631	.315	.472	.953	.016	.253	.736	.409	.536	.710	.049
Motor Planning and Coordination	Purdue Pegboard	Corr.	.062	−.179	.198	−.031	.038	−.302	.055	−.158	.169	−.011	−.018	−.215
		Sig.	.685	.240	.193	.841	.804	.044	.722	.306	.272	.944	.906	.161

Note. TST = total sleep time; SE = sleep efficiency; WASO = wake after sleep onset; NOA = number of awakenings; SFI = sleep fragmentation index. Sig. = Significance (2-tailed); Corr. = Correlation. Bolded text, p <.05.

3.4. Cognitively Unaffected versus MCI-DS

In independent sample t-tests, participants with MCI-DS had significantly higher mean actigraph LOA and lower mean SE than cognitively unaffected participants (Table 1). There were no other significant differences in actigraph sleep indices between the clinical AD status groups. Partial correlations as shown in Table 5, adjusting for overall cognitive ability (PPVT-4), were used to examine the association between actigraph sleep measures and AD-related cognitive domains in the cognitively unaffected subset (N=40). Greater mean LOA was associated with poorer performance in Cued Recall Intrusions (r = .371, p = .022), Cat and Dog Stroop Switch Time (r = .341, p = .036) and Free and Cued Recall (r =−.332, p =.042). Greater actigraph SFI was associated with poorer performance in Cued Recall Intrusions (r = .349, p =.032) and Free and Cued Recall (r = −.329, p =.044). In the cognitively unaffected individuals, when additionally adjusting for chronological age, the associations between actigraph LOA and SFI and Free and Cued Recall and Cued Recall Intrusions remained significant (Table 5).

Table 5.

Association between AD-related cognitive domains and actigraph sleep indices in the cognitively unaffected subset (N=40)

			Adjusting for Mental Age						Adjusting for Mental Age and Chronological Age
			TST	WASO	SE	SFI	NOA	LOA	TST	WASO	SE	SFI	NOA	LOA
Working Memory	Free and Cued	Corr.	.127	−.145	.241	−.329	−.111	−.332	.136	−.149	.247	−.335	−.117	−.333
		Sig.	.449	.386	.145	.044	.506	.042	.422	.379	.140	.043	.489	.044
	Cued Recall Intrusions	Corr.	−.071	.128	−.179	.349	.048	.371	−.089	.137	−.191	.361	.059	.359
		Sig.	.672	.443	.282	.032	.776	.022	.602	.419	.257	.028	.728	.029
	Rivermead	Corr.	−.090	.151	−.174	−.034	.100	−.070	−.102	.157	−.181	−.028	.108	−.094
		Sig.	.592	.364	.297	.841	.550	.677	.549	.353	.283	.870	.525	.581
	Digit Span Forward	Corr.	.130	−.025	.133	−.122	−.137	−.137	.150	−.033	.146	−.133	−.151	−.110
		Sig.	.437	.883	.425	.465	.412	.412	.376	.844	.389	.431	.374	.517
Executive Functioning	Digit Span Backward	Corr.	.023	.144	−.029	−.044	.216	−.020	.039	.138	−.020	−.053	.208	.009
		Sig.	.891	.388	.862	.792	.193	.905	.819	.416	.907	.754	.217	.959
	Cat and Dog	Corr.	−.182	.154	−.177	.164	.112	.341	−.216	.171	−.199	.184	.135	.306
		Sig.	.275	.357	.289	.324	.502	.036	.198	.313	.239	.275	.426	.066
Motor Planning and Coordination	Purdue Pegboard	Corr.	−.013	−.107	.095	.030	.008	−.300	.035	−.135	.130	.004	−.025	−.234
		Sig.	.937	.523	.570	.859	.963	.067	.839	.425	.444	.982	.885	.162

Note. TST = total sleep time; SE = sleep efficiency; WASO = wake after sleep onset; NOA = number of awakenings; SFI = sleep fragmentation index. Sig. = Significance (2-tailed); Corr. = Correlation. Bolded text, p <.05

4. Discussion

In non-DS populations, there is growing evidence (Bubu et al., 2017; Ju et al.,2013; Spira et al., 2013; Sprecher et al., 2015; Sprecher et al., 2017) of an association between sleep problems and β-amyloid accumulation prior to the development of clinical AD dementia. The goal of the current study was to examine the association between sleep, PiB PET β-amyloid accumulation, and performance on AD-related cognitive measures of memory, executive functioning, and motor planning and coordination in a sample of non-demented adults with DS. Consistent with previous reports of sleep in the DS population (Edgin et al., 2015; Esbensen et al., 2018; Stores, 2019; Trois et al., 2009), this cohort of adults with DS evidenced poor sleep, with mean sleep efficiency ranging from 58% to 88% (M = 75.38, SD = 7.65) and an average of 23 awakenings per night. Average length of nighttime awakenings was significantly associated with higher striatal β-amyloid burden, a region known (Handen et al., 2012; Lao et al., 2016; Lao et al., 2017) to evidence early AD-related β-amyloid accumulation in DS. These associations occurred in adults with DS in the earliest phases of AD, without clinical dementia, and align with findings from other at-risk populations (Bubu et al., 2017; Ju et al.,2013; Spira et al., 2013; Sprecher et al., 2015; Sprecher et al., 2017) demonstrating that sleep disruptions may be involved early on in the AD pathophysiological cascade. The association between length of nighttime awakenings and striatal β-amyloid was modest (r =.323) and became non-significant when controlling for chronological age. It is important to note, however, that chronological age was strongly associated with global (r = .609, p <.001) and striatal (r=.773, p <.001) β-amyloid burden in the current sample. This tight linkage makes disentangling the effect of normative aging from the effects of early AD-related pathophysiological changes challenging, particularly as these processes largely go hand-in-hand in the DS population.

After adjustment for overall cognitive ability, non-demented adults with DS with longer average length of nighttime awakenings performed worse on measures of episodic memory (Cued Recall Intrusions), executive functioning (Cat/Dog Stroop Switch Time), and motor planning and coordination (Purdue Pegboard). For the most part, these associations remained after adjusting for chronological age, and a similar pattern of associations were observed after excluding individuals with MCI-DS. Namely, when examining sleep in the cognitively unaffected subjects, measures of sleep disruption like length of nighttime awakenings and sleep fragmentation were significantly related to poorer performance in working memory and executive functioning tasks. These results are consistent with previous findings (Edgin et al., 2015) indicating that disrupted sleep may negatively affect cognitive performance in individuals with DS. Moreover, these findings underscore the significant role of sleep on cognitive performance in the early AD stages, prior to the onset of MCI or dementia in adults with DS.

Due to the observational and cross-sectional design, we were unable to determine whether disrupted sleep drives β-amyloid accumulation or vice versa. Animal studies suggest the relationship between sleep and β-amyloid may be bidirectional. Research on APP transgenic mice, which similar to individuals with DS all develop β-amyloid plaques, provides evidence for a destructive feedback loop whereby sleep deprivation increases β-amyloid deposition and prolonged β-amyloid aggregation disrupts the sleep-wake cycle (Kang et al., 2009; Roh et al., 2012). Congruently, we showed that MCI-DS individuals, all with increased β-amyloid burden, had significantly higher average length of nighttime awakenings compared to cognitively unaffected individuals with largely lower levels of β-amyloid. These results raise the possibility that increased sleep disruptions may occur with increasing levels of β-amyloid; however, the current study cannot evaluate causality or directionality. Future longitudinal studies are needed to tease apart the time-ordered pathways between disrupted sleep, β-amyloid accumulation, and cognitive decline in DS.

Strengths of the study include the assessment of a well-characterized cohort of adults with DS and the use of directly-administered cognitive measures, PiB PET neuroimaging, and actigraph accelerometry. There were also several study limitations. While approximately half of the participants had concurrent imaging, cognitive, and sleep assessments, the time interval between the imaging and cognitive assessments and the sleep assessment ranged from 0 to 1.5 years. There is a possibility that changes in sleep (e.g., developed new sleep problems or began treatment for existing problems) occurred during this period. It is also possible that some adults with DS in the cognitively unaffected group had begun on a trajectory of cognitive decline, but these changes were not detected in the case consensus process. The study is also limited in that objective measures of sleep disordered breathing were not obtained. Given the high prevalence of obstructive sleep apnea in DS (Trois et al., 2009), the association between length of nighttime awakenings, β-amyloid accumulation, and cognitive performance may be driven by or exacerbated by the effect of nocturnal hypoxia (e.g., Sharma et al., 2018). Thus, methods that collect information on oxygen levels during sleep should be included in future studies of sleep in DS. Similarly, given the evidence of impaired REM sleep in non-DS populations (Pase et al., 2017), future studies incorporating polysomnography would allow for examination of the stages of sleep most disrupted in AD in the DS population. Subsequent DS research should also evaluate patterns of circadian rest-activity rhythms, which are posited to be involved in β-amyloid clearance, inflammation, and cerebral blood flow (Homolak et al, 2018). These additional methodologies and information are critical for identifying the causal mechanisms driving the association between sleep disruptions and AD and important for identifying potential therapeutic or intervention targets.

To our knowledge, this study is the first to examine the role of sleep in AD-related β-amyloid and cognitive functioning in adults with DS. Our results may have significant public health implications. Nearly all individuals with DS develop AD, and one-third to one-half of adults with DS (Stores, 2019; Trois et al., 2009) experience disrupted sleep. As sleep disturbance can be treated, interventions to improve sleep among individuals with DS may help slow the onset of β-amyloid accumulation and AD-related cognitive decline. This could have a significant effect on both the quality of life of individuals with DS and their families as well as on the considerable cost of healthcare associated with AD.

In summary, sleep disruption, specifically length of awakenings at night, was associated with elevated striatal β-amyloid and poor memory, executive functioning, and motor planning and coordination cross- sectionally in adults with DS in the preclinical stages of AD. Relatedly, adults with DS and MCI-DS, all of whom evidenced β-amyloid accumulation, showed a longer average length of nighttime awakenings compared to cognitively unaffected adults with DS. These findings implicate sleep as a potential modifiable risk factor for AD that could be targeted at early life stages in DS. If sleep problems contribute to and/or exacerbate the early stages of AD pathophysiology, efforts to reduce sleep disruptions may allow adults with DS to delay the onset of clinical AD dementia.

Highlights.

• Adults with Down syndrome have an increased risk for both sleep problems and Alzheimer’s disease

• Disrupted sleep was associated with β-amyloid in non-demented Down syndrome adults

• Disrupted sleep was also associated with worse cognitive performance

• Sleep may be implicated in the preclinical stages of AD in Down syndrome