Disrupted rest-activity rhythms and cerebral small vessel disease pathology in older adults

Abstract

Background and Purpose:

The pathogenesis of cerebral small vessel disease (CSVD) remains incompletely understood. The relationship between circadian rhythm disturbances and histopathological measures of CSVD has not been studied. We hypothesized that disrupted circadian rest-activity rhythms would be associated with a higher burden of CSVD pathology.

Methods:

We studied 561 community-dwelling older adults (mean age at death = 91.2, 27.4% male) from the Rush Memory and Aging Project. We used actigraphy to quantify several measures of 24-hour rest activity rhythmicity, including interdaily stability (IS), intradaily variability (IV), and amplitude, and used ordinal logistic regression models to relate these measures to the severity of cerebral arteriolosclerosis, atherosclerosis, macroinfarcts, and microinfarcts, assessed at autopsy.

Results:

Lower IS was associated with a higher burden of arteriolosclerosis, higher IV was associated with a higher burden of atherosclerosis and subcortical infarcts, and lower amplitude was associated with a higher burden of arteriosclerosis, atherosclerosis and subcortical macroinfarcts. Moreover, the associations between IS and arteriolosclerosis and IV and subcortical infarcts were independent of cardiovascular risk factors, sleep fragmentation, and medical comorbidities.

Conclusions:

Disrupted rest activity rhythms are associated with a greater burden of cerebral small vessel disease in older adults.

Introduction

Cerebral small vessel disease (CSVD), damage to the small vessels of the brain, is a key pathological correlate of cognitive impairment in older adults¹. However, its pathogenesis is poorly understood, and classical vascular risk factors explain only some of the variance in the burden of pathology². In human experiments, circadian rhythm disruption causes physiological changes that may predispose to CSVD³. We had previously shown that unstable rest-activity rhythms, markers of circadian rhythmicity, are associated with the metabolic syndrome and its components⁴. Moreover, unstable and fragmented 24-hour rest-activity rhythms are associated with magnetic resonance imaging markers of CSVD, such as cerebral microbleeds and white matter lesions⁵. However, there is few data concerning whether unstable, fragmented, or low amplitude rest-activity rhythms are associated with histological measures of CSVD. Building on our prior work showing that nocturnal sleep fragmentation is associated with histopathological CSVD markers⁶, we hypothesized that unstable, fragmented, or low amplitude 24-hour rest-activity rhythms, independent of sleep fragmentation, would be associated with a higher burden of CSVD histopathology in community dwelling older adults.

Methods

Please see https://www.ahajournals.org/journal/str for a full description of the methods. The Rush Memory and Aging Project was approved by the Institutional Review Board of Rush University Medical Center.

All data is available through the RADC Research Resource Sharing Hub (https://www.radc.rush.edu/).

Study participants

We studied 561 participants from the Rush Memory and Aging Project, a community-based cohort study of older individuals⁷. The Rush Memory and Aging Project was approved by the Institutional Review Board of Rush University Medical Center. The study was conducted in accordance with the latest version of the Declaration of Helsinki, and all participants provided written informed consent of the full study protocol, including a signed anatomic gift act for brain donation and a repository consent for data sharing. We included all participants with at least one actigraphic recording and available post-mortem measures of CSVD.

24-Hour Rest-Activity Rhythms

Participants wore an actigraph (Actical, Phillips Respironics, Bend, OR) biennially for 10 consecutive days on their non-dominant wrist. After exclusion of days containing periods of non-wear, as described in the supplemental methods, the first seven complete days of data were used to calculate three primary metrics of diurnal rhythmicity: 1) interdaily stability (IS)⁸, which measures the stability of activity patterns from day to day (0=no similarity; 1=all days identical); 2) intradaily variability (IV)⁸, which measures the fragmentation of activity rhythms (higher values represent greater fragmentation of rhythms); and 3) the amplitude of rest activity rhythms, or the height of the cosine curve fit of log-transformed activity counts. All three measures have been reported to be associated with MRI markers of CSVD^{5, 9}. For participants with more than one cycle of actigraphy, the mean of all measurements was used, as this best represents the cumulative exposure to disrupted circadian rhythmicity in the years leading to death.

Consensus Diagnosis of Dementia

At the time of death, all available clinical data were reviewed by a neurologist with expertise in dementia, and a summary diagnostic opinion was rendered regarding the most likely clinical diagnosis at the time of death, according to National Institutes of Neurological and Communication Disorders and Stroke – Alzheimer’s Disease and Related Disorders Association criteria¹⁰. Summary diagnoses were made blinded to all post-mortem data.

Neuropathology

After death, brains were removed, hemisected, and cut into 1cm coronal sections, with 1 hemisphere fixed in 4% paraformaldehyde. Atherosclerosis was evaluated by visual inspection of the Circle of Willis and graded on a 4-point semi-quantitative (none, mild, moderate, severe) scale¹¹. Arteriolosclerosis was evaluated by histological assessment of the small vessels of the anterior basal ganglia and graded on a 4-point semiquantitative scale (none, mild, moderate, severe)¹¹. Macroscopic cerebral infarcts were identified as cavitating or incomplete lesions based on visual inspection of the slabs and confirmed by histological examination¹¹. Microinfarcts were identified by histological examination¹¹. The number of cerebral infarcts were converted to an ordinal outcome of either none, one, or two or more. All infarcts were classified as either cortical or subcortical.

Covariates

Post-mortem interval was defined as the number of hours from death to autopsy. As described in the supplemental methods, we also computed k_RA, a metric of sleep fragmentation, from the actigraphic data. Briefly, k_RA represents the probability of arousal (as indicated by movement) per unit time following a period of sustained rest. Age, sex, level of education, and vascular risk factors, the metabolic syndrome (diabetes, dyslipidemia, hypertension, and obesity), medical comorbidities (cardiovascular disease, cancer, and thyroid disease), actigraphy-derived mean total daily activity were assessed as described in the online supplement.

Statistical analysis

We used multivariable ordinal logistic regression models to test the association between IS, IV, and amplitude and the burden of arteriolosclerosis, atherosclerosis, cortical and subcortical macro- and microinfarcts, adjusting for age at death, sex, and years of education. Multiple comparisons were accounted for by computing a false discovery rate. We then considered additional models adjusting for potential confounders, including post-mortem interval, clinical history of stroke, presence of metabolic syndrome, number of vascular risk factors, physical activity, and sleep fragmentation. Finally, we used logistic regression models to test the associations between IS, IV, amplitude, arteriolosclerosis, atherosclerosis, and subcortical macroscopic infarcts, and the odds of having a consensus diagnosis of dementia.

Results

Please see https://www.ahajournals.org/journal/str for supplemental tables and figures. Data from 561 participants were included in these analyses. Their characteristics are described in Table 1. 27.4% of participants were male and participants had a mean of 14.8 years of education. The mean age at death for these participants was 91.2 years, and the last actigraphic recording was obtained a mean of 2.5 years prior to death. These participants had a mean IS of 0.52, mean IV of 1.33, and mean amplitude of 4.76. 42% of the participants had a consensus clinical diagnosis of dementia. Our CVSD outcome measures were generally weakly correlated with each other (Supplemental Table I). However, our circadian rhythm measures IS, IV, and amplitude were moderately to strongly correlated with each other (Supplemental Table II).

Table 1:

Study population characteristics

Variable		Mean [SD] or n (%)
N (% Male)		561 (27.5%)
Education (years)		14.8 [2.8]
Age at death		91.2 [6.1]
Age at first actigraphy		85.2 [5.8]
Age at last actigraphy		88.6 [6.1]
Number with a clinical diagnosis of dementia		233 (42%)
Number of Actigraphic Recordings		3.6 [2.6]
Years since last actigraphy		2.6 [2.4]
Interdaily Stability (IS)		0.52 [0.12]
Intradaily Variability (IV)		1.33 [0.25]
Amplitude		4.76 [2.5]
BMI (kg/m2)		26.2 [4.9]
Systolic Blood Pressure (mmHg)		133 [16.8]
Diastolic Blood Pressure (mmHg)		75 [9.5]
Presence of any risk factor		547 (97.5%)
	Smoking	216 (38.5%)
	Diabetes	128 (22.8%)
	Hypertension	535 (93.4%)
Metabolic syndrome		159 (28.3%)
Medical comorbidities		439 (78.3%)
	Cardiovascular Disease	198 (35.3%)
	Thyroid Disease	146 (26.0%)
	Cancer	223 (39.8%)
Arteriolosclerosis
None		192 (34.2%)
Mild		205 (36.5%)
Moderate		130 (23.2%)
Severe		34 (6.0%)
Atherosclerosis
None		134 (23.9%)
Mild		294 (52.4%)
Moderate		109 (19.4%)
Severe		24 (4.2%)
Cortical Macroinfarcts
None		473 (84.3%)
1		61 (10.9%)
2+		27 (4.8%)
Subcortical Macroinfarcts
None		396 (70.6%)
1		103 (18.4%)
2+		62 (11.1%)
Cortical Microinfarcts
None		473 (76.8%)
1		87 (15.6%)
2+		43 (7.7%)
Subcortical Microinfarcts
None		478 (85.2%)
1		62 (11.1%)
2+		21 (3.7%)

In models controlling for age, sex, and years of education, each standard deviation (SD) lower IS, representing more unstable 24-hour rest-activity rhythms, was associated with 19% higher odds of having a higher grade of arteriolosclerosis (i.e. severe vs. moderate or lower; moderate vs. mild or none; mild vs. none), equivalent in effect to 4.1 years of aging (Table 2; Supplemental Figure I). In contrast, IS was not associated with atherosclerosis (Table 2; Supplemental Figure II). There was an association between lower IS and a higher burden of subcortical macroinfarcts, which did not attain significance (Supplemental Figure III), but not cortical macroinfarcts, or any microinfarcts (Table 2). In models augmented with interaction terms for sex, we did not find statistically significant differences in effect estimates between men and women (p>0.05 for all).

Table 2:

Ordinal logistic regression models relating actigraphically derived measures of circadian rhythmicity (interdaily stability, intradaily variability, amplitude), and cerebrovascular pathology (arteriolsclerosis, atherosclerosis, subcortical and cortical macro- and microinfarcts), adjusted for age at death, sex, and years of education.

	Arteriolosclerosis	Atherosclerosis	Subcortical infarct	Cortical infarct	Subcortical microinfarct	Cortical Microinfarct
IS	1.19 [1.02–1.39]; p = 0.03 (q = 0.10)	1.14 [0.97–1.33]; p = 0.11 (q = 0.19)	1.20 [1.00–1.43]; p = 0.05 (q = 0.14)	0.89 [0.71–1.12]; p = 0.34 (q = 0.40)	1.20 [ 0.95–1.52]; p = 0.12 (q = 0.19)	1.11 [0.92–1.35]; p = 0.29 (q = 0.37)
IV	1.16 [0.99–1.36]; p = 0.069 (q = 0.15)	1.25 [1.06–1.48]; p = 0.009 (q = 0.054)	1.54 [1.28–1.87]; p = 0.000 (q < 0.001)	1.14 [0.90–1.45]; p = 0.27 (q = 0.37)	1.12 [0.88–1.44]; p = 0.36 (q = 0.40)	0.87 [0.71–1.07]; p = 0.19 (q = 0.28)
Amplitude	1.19 [1.02–1.40]; p = 0.03 (q= 0.10)	1.18 [1.00–1.39]; p = 0.05 (q= 0.14)	1.35 [1.12–1.62]; p = 0.001 (q= 0.009)	1.09 [0.87–1.38]; p = 0.45 (q = 0.47)	1.24 [0.98–1.57]; p = 0.08 (q = 0.15)	0.95 [0.78–1.16]; p = 0.60 (q = 0.60)

Each SD of higher IV, representing more fragmented 24-hour rest-activity rhythms, was associated with 25% higher odds of having more severe atherosclerosis and 54% higher odds of having more subcortical macroinfarcts (Table 2, Supplemental Figures IV–V). There was also an association between higher IV and a higher burden of arteriolosclerosis, which did not attain significance (Supplemental Figure VI). Moreover, each SD lower amplitude of circadian rhythmicity was also associated with higher odds of having a higher grade of arteriolosclerosis, more subcortical macroinfarcts, and more severe atherosclerosis (Table 2; Supplemental Figures VII–IX). There were no significant associations between IV, amplitude, and cortical macroinfarcts or microinfarcts (Table 2).

The association between the stability of rest-activity rhythms and arteriolosclerosis remained significant when controlling for the post-mortem interval, actigraphically-quantified sleep fragmentation, daily physical activity, vascular risk factors and diseases, systolic blood pressure, use of antihypertensive and cardiac medications, presence of the metabolic syndrome, thyroid disease, and cancer (Supplemental Table III). Similarly, adjusting for these confounders did not attenuate the effect of greater IV on the number of subcortical infarcts (Supplemental Table IV). In a sensitivity analysis including only participants with the latest actigraphic data obtained within 4 years of death (N = 444), these associations remained significant (Supplemental Table V).

In logistic regression models adjusted for age, sex, and years of education, lower IS, amplitude, and higher IV were associated with higher odds of having a consensus clinical diagnosis of dementia (Supplementary Table VI). Moreover, the presence of one or more subcortical infarcts (OR = 1.54; 95% CI = 1.06–2.23), presence of mild or greater arteriolosclerosis (OR = 1.61; 95% CI = 1.11–2.33) and presence of moderate or greater atherosclerosis (OR = 1.70; 95% CI = 1.14–2.52) were also associated with higher odds of having a consensus diagnosis of dementia. Inclusion of terms for arteriolosclerosis, atherosclerosis, and subcortical infarcts in models relating IS, IV, and amplitude to the odds of dementia somewhat attenuated these associations, compatible with these CSVD pathologies mediating a modest portion of the association between disrupted rest-activity rhythms and a clinical diagnosis of dementia (Supplemental Table VI).

Discussion

In this study of 561 community-dwelling older adults, more disrupted 24-hour rest-activity rhythms, objectively quantified by continuous actigraphic recordings, were associated with a higher burden of arteriolosclerosis and subcortical infarcts at death. These associations were independent of vascular risk factors and medical co-morbidities.

The rest-activity rhythms in this study were both more unstable and more fragmented than previous studies that examined associations between rest-activity rhythms and imaging markers of SVD in older adults⁵. Age has been shown to be associated with more fragmented^{12, 13} and lower amplitude¹³ rest activity rhythms, and our analytic sample was considerably older (mean age 85.2 years at the time of first actigraphy) than this prior study (mean age 59.2 years at the time of actigraphy).

Associations between circadian misalignment in shift workers and cardiovascular risk factors and the metabolic syndrome have been reported¹⁴. Moreover, unstable rest-activity rhythms were reported to be associated with MRI markers of CSVD⁵. However, what this represents at the histopathological level was not known. Our data suggest that these imaging associations are reflective of key CSVD neuropathologies including arteriolosclerosis and subcortical infarcts.

We hypothesize that lower IS, higher IV, and decreased amplitude of rest-activity rhythms may lead to physiological changes that contribute to CSVD pathology. This is consistent with a body of human experimental studies showing that forced desynchronization of behavioural and circadian rhythms result in increased blood pressure, decreased vagal tone, increased inflammation, and glucose intolerance³. Other interpretations are possible but less likely. Ischemic injury to circadian circuits could in theory disrupt circadian rhythms; however, the hypothalamus is not prone to vascular damage, and hypothalamic ischemia would likely lead to alterations of other autonomic functions. It is also possible that underlying risk factors may be contributing to both disrupted rest-activity rhythms and CSVD. However, the association between unstable rhythms and arteriolosclerosis was independent of known vascular risk factors.

While more disrupted rest-activity rhythms were associated with arteriolosclerosis, atherosclerosis, and subcortical macrosocpic infarcts, they were not associated with cortical macroscopic infarcts or cortical or subcortical microscopic infarcts. There is evidence that different SVD markers may have different underlying pathophysiological bases. For instance, cortical infarcts are more likely to be associated with atrial fibrillation and other cardioembolic conditions than subcortical infarcts¹⁵, and in contrast with cerebral macroinfarcts, cerebral microinfarcts are associated with amyloid angiopathy¹⁶ but not with diabetes and hypertension¹⁷. Further studies are needed to elucidate exactly which specific pathophysiological processes mediate the link between diurnal rest activity rhythms and specific CVSD pathologies.

There are several limitations. First, an observational cross-sectional study does not allow definitive determination of temporal sequence or causal direction. Second, actigraphy, while well-tolerated and objectively measures 24-hour activity rhythms, cannot distinguish dysfunction of the circadian clock from masking effects by voluntary behaviour and environmental factors. Third, participants were predominantly older, female, and of European ancestry; thus, these findings need to be confirmed in other populations. Fourth, there were relatively few participants with severe CSVD, which may have limited the power to draw conclusions about individuals with severe CSVD. Lastly, while we adjusted for overt clinical diagnoses of a number of potential confounders including cardiac disease, peripheral vascular disease, thyroid disease, cancer, and others, we did not have quantitative markers of subclinical disease, which is a limitation.

This study has several strengths. We used objective multi-day measures of rest-activity rhythms, avoiding recall-bias and taking into consideration day-to-day variation. Moreover, we used gold-standard histological outcomes. Finally, study participants were well-characterized, including objective measures of vascular risk.

Notwithstanding these limitations, we conclude that in older community dwelling adults, disrupted 24-hour rest-activity rhythms are associated with a greater burden of arteriolosclerosis and subcortical infarcts and may be a marker of, or a contributor to them. Further longitudinal observational studies with repeated measurement of rest-activity rhythms and in vivo CSVD assessments are required to establish the temporal relationship between circadian disruption and CSVD. Additionally, experimental disruption of circadian rhythms in animals will help to elucidate the biological mechanisms underlying this association. Eventually, clinical trials of circadian interventions will be needed before the clinical significance of circadian disruption as a contributor to human CSVD can be established.

Non-standard Abbreviations and Acronyms

CSVD

Cerebral small vessel disease

IS

Interdaily stability

IV

Intradaily variability