Basal Cortisol Levels Are Increased in Patients with Mild Cognitive Impairment: Role of Insomnia and Short Sleep Duration

Abstract

Background:

Mild cognitive impairment (MCI) is frequent in elderly and a risk factor for dementia. Both insomnia and increased cortisol levels are risk factors for MCI.

Objective:

We examined cross-sectionally whether increased cortisol levels are associated with short sleep duration (SSD) and/or the insomnia short sleep duration (ISS) phenotype, in elderly with MCI.

Methods:

One hundred twenty-four participants with MCI and 84 cognitively non-impaired controls (CNI)≥60 years underwent medical history, physical examination, neuropsychiatric evaluation, neuropsychological testing, 3-day actigraphy, assessment of subjective insomnia symptoms, and a single morning plasma cortisol level. The short sleep phenotypes were defined by sleep efficiency below the median of the entire sample (i.e.,≤81%) with at least one insomnia symptom (ISS) or without (SSD). ANOVA models were used to compare the various sleep phenotypes to those who did not present either short sleep or insomnia symptoms [non-insomnia (NI)].

Results:

MCI participants had higher cortisol levels compared to the CNI group (p = 0.009). MCI participants with insomnia (n = 44) or SSD (n = 38) had higher cortisol levels compared to the NI group (n = 42; p = 0.014 and p = 0.045, respectively). Furthermore, MCI participants with ISS phenotype but not those with insomnia with normal sleep duration had higher cortisol levels compared to NI (p = 0.011 and p = 0.4, respectively). Both linear trend analyses showed that cortisol reached the highest levels in the ISS phenotype.

Conclusion:

The ISS and SSD phenotypes are associated with increased cortisol levels in elderly with MCI. Improving sleep quality and duration and decreasing cortisol levels may delay further cognitive decline.

Keywords

Cortisol insomnia mild cognitive impairment sleep duration

INTRODUCTION

Aging of the population is a characteristic of modern western societies [1], and, based on data from 196 countries of different regional and socioeconomic background, Greece has the 6th highest prevalence (26.9% of the general population) of people aged 60 years or more [2].

The prevalence of mild cognitive impairment (MCI), a condition of memory difficulties and executive impairment but without an adverse effect on daily function, is estimated to be between 16% and 32% among those older than 60 and is associated with high risk for dementia [3 –6]. Several lifestyle and behavioral factors, such as obesity, sleep, physical activity, depression, loneliness, and stress, have been reported to be associated with the progression of MCI into dementia [7].

Elevated basal cortisol levels have been linked to dysregulation of the hypothalamic-pituitary-adrenal (HPA) axis, which is related to excessive exposure to physical or psychosocial stressors, and aging [8 –13]. Furthermore, increased basal cortisol levels have been associated with a reduced volume of the hippocampus and prefrontal cortex in older adults, crucial brain anatomic structures for memory and executive functions [8 , 14–17]. Several studies in older adults have revealed negative associations between basal cortisol levels, memory performance, and executive functions [18 –20].

Previous studies have compared peripheral or central levels of cortisol between patients with MCI and cognitively non-impaired controls and the results are inconsistent [21 –25]. Based on two recent extensive reviews and meta-analyses the authors concluded that cerebrospinal fluid (CSF) but not plasma or saliva levels are moderately elevated in patients with MCI [26], whereas plasma cortisol levels were reported to be elevated only in patients with MCI due to Alzheimer’s disease, a more severe type of pre-dementia MCI [27].

Sleep disturbances, either objective or subjective, are common in the elderly with as many as 50% of persons over 65 years reporting a chronic sleep complaint [28]. Moreover, further significant changes in terms of sleep quality occur even in early stages of cognitive disorders [29]. Previous studies based primarily on self-reported and fewer more recent based on objective measures, have indicated that short sleep duration is associated with adverse effects on cardiovascular function [30 –33] and memory [34, 35]. Additionally, it has been reported that insomnia-type symptoms in community-dwelling elderly are associated with cognitive decline [36 –40]. Furthermore, insomnia with objective short sleep duration (ISS) has been proposed as a novel more severe phenotype associated with activation of the stress system, particularly the HPA axis and significant impact on health such as increased hypertension, impaired heart rate variability, diabetes type II, neurocognitive impairment, and mortality [30 , 41–44]. To our knowledge, the association of the ISS phenotype with peripheral levels of cortisol in patients with MCI has not been examined.

The aim of the present study that included a large sample of MCI participants and controls was to examine cross-sectionally 1) whether the plasma basal levels of cortisol are higher in participants with MCI compared to cognitively non-impaired controls and 2) whether the higher levels of cortisol are associated with the short sleep duration (SSD) and ISS phenotype. We hypothesized that MCI participants will demonstrate higher basal plasma cortisol levels compared to cognitively non-impaired (CNI) controls and that the increased levels of cortisol are primarily associated with the SSD and ISS phenotypes.

MATERIALS AND METHODS

Study design

The current sample included participants of the Cretan Aging Cohort (CAC), a cross-sectional study of community dwelling elders, recruited from the area of Heraklion, Crete, Greece, focusing on the prevalence and risk factors associated with cognitive impairment [6]. The Cretan Aging Cohort study was conducted in two phases, as described in detail in our previous publication [6] and illustrated in Fig. 1. The study was approved by the Bioethics Committee of the University Hospital of Heraklion, Crete (Protocol Number: 13541, 20-11-2010).

Fig. 1

Study flowchart. MCI, mild cognitive impairment; CNI, cognitively non-impaired; MMSE, Mini-Mental State Examination.

Phase I

Eligible participants were those aged≥60 years who visited selected Primary Health Care facilities in rural and urban areas of the Heraklion district for any reason. All consenting individuals (n = 3,200) completed an interview with a trained nurse, using a structured questionnaire including sociodemographic information, lifestyle, and anthropometric measurements. Additionally, physical and mental health problems, as well as medication use were assessed and confirmed by their primary care physician [45]. Evaluation of cognitive function was conducted using the Greek version of the Mini-Mental State Examination (MMSE) test [46].

Participants with crucial missing data (MMSE, age) were excluded. The final study sample consisted of 3,140 persons (57.0% women), aged 73.7±7.8 (range: 60–100) years who had completed an average of 5.8±3.3 (0–18) years of formal education (82.7% and 23.9% with≤6 years and≤4 years of formal education, respectively), mainly residing in rural areas (86.2%).

Phase II

All participants who scored < 24 points on MMSE (n = 636) in Phase I were invited to receive a comprehensive neuropsychological and neuropsychiatric evaluation (Phase II) and 344 agreed to participate (response rate: 54.1%). Participants with low MMSE who were not evaluated for any reason (n = 292; refused, could not be located, or passed away) did not differ from the 636 subjects referred, in terms of age, gender, and body mass index (BMI). A comparison group, the size of which was set to approximately 30% of the low MMSE group, was formed by applying a proportional stratification procedure using residence (geographic region within the Heraklion prefecture), age, and gender as the stratification variables to the remaining 2,504 persons who scored≥24 points on the MMSE. Using this procedure, 181 persons were invited to take part in phase II, and 161 agreed to participate.

During Phase II, assessment involved administration of an extensive questionnaire (modified from the one employed in the HELIAD study [47]), including determinants of health status in a semi-structured interview. Neuropsychiatric assessment was conducted by a team of certified neurologists, psychiatrists, and internists. Neuropsychological assessment was performed by trained neuropsychologists using a test battery described in detail elsewhere [45]. Diagnosis of any type of MCI was based on modified Petersen criteria (IWG-1) [48] and on a consensus decision between two or more clinicians who took into account results from the comprehensive neuropsychiatric and neuropsychological evaluation. Diagnosis of MCI further required that cognitive deficits could not be accounted for by clinically significant mood or anxiety disorder. To be included in the MCI group, patients had to have age- and education-adjusted z scores < –1.5 on indices derived from at least two tests within a given cognitive domain (episodic memory, language, attention/executive) and demonstrate intact levels of every-day functionality according to a comprehensive informant Instrumental Activities of Daily Living scale (see Fig. 1).

Participants

A sub-sample of 208 Phase II participants had both 24 h actigraphy and morning serum cortisol measurements: 124 met clinical and neuropsychological criteria for MCI whereas 84 persons were included in the cognitively non-impaired comparison group (CNI; 74.86±7.1 years old) (see Fig. 1).

Measures

Insomnia symptoms and objective sleep duration

Self-reported sleep and presence of all sleep disorders during Phase II were evaluated using a modified standardized 53-item questionnaire developed by the Penn State sleep group and used extensively in this group’s publications since 1998 [49]. Each question was rated on a 0 (“no/never symptoms”), 1 (“mild/ occasionally”), 2 (“moderate/ often”), and 3 (“severe/ always”) scale [49]. The presence of “insomnia symptoms” was defined by the report of any insomnia symptom based on a “yes” answer on the questions “Do you have difficulty falling asleep”, “Do you have difficulty staying asleep”, and “Do you wake up in the morning earlier than desired” often or always. Questions on weekly frequency, duration, or interference on daily function were not included.

Objective measurement of sleep was recorded for three consecutive nights through actigraphy device (Actigraph, GT3XP model, Pensacola, FL, USA). The device was placed on the non-dominant wrist and the participant was instructed to fill out sleep diaries reporting “bedtime”/“out of bedtime”, as well as times the actigraph was removed on a daily basis. Night was defined based on the questions “what time did you go to bed” and “what time did you get out of bed” filled by the participant or caregiver in the sleep diary. Analysis of actigraphy data was conducted using the ActiLife 6 software (Actilife v6.9.5 LLC, Pensacola, FL, USA) and involved computation of sleep latency, wake time after sleep onset, total sleep time, and sleep efficiency. Actigraphy recordings were conducted during weekdays for all participants. The start time of the 24-hour period was set at 11:00 a.m. on the day that the actigraph was applied on the participant’s hand by our staff. Participants with actigraphic recording of less than 3 days or daily night total sleep time (TST)≤3 hours were excluded from the analysis. The presence of “short sleep duration” was defined by an actigraphy-measured sleep efficiency below the median of the entire sample (i.e.,≤81%). The ISS phenotype was defined by the presence of at least one subjective insomnia symptom and short sleep duration.

Cortisol measurement

Single, morning, blood samples were collected between 10:00 a.m. and 12:00 p.m., transferred to EDTA-containing tubes (3 per patient) and refrigerated until centrifugation (within 3 h) for the plasma isolation which was kept in deep freeze (–80°C). Plasma cortisol was measured by ELISA technique (Cusabio Human cortisol ELISA kit, cfb-e05111h, Cusabio Technology LLD, Texas, USA). The inter- and intra-assay coefficients of variation were 23.1% and 16.1%, respectively. The lower detection limits for cortisol were 0.02 ngr/ml.

Other variables

Sociodemographic variables such as gender, age, education level, and living alone were assessed. Also, prescription of any type of psychotropic medication (i.e., antipsychotics, antidepressants, anxiolytics, and benzodiazepines) was recorded. Furthermore, regular physical activity was based on participant responses to a single question “How many days during the 7 days have you walked for more than 10 minutes continuously”. Lack of physical activity was defined if participants had less than 3 days of activity during the week.

Statistical analysis

Data are presented as the mean±SD for continuous variables and frequency and percent for categorical variables. The participants’ demographic and clinical characteristics were summarized and compared by using T-test, chi-square test, or ANOVA tests as appropriate. The effect of insomnia phenotype among MCI patients on log-transformed cortisol levels was assessed in a step-wise fashion: initially, the effect of short sleep duration (without insomnia), and insomnia (regardless of objective sleep duration) was examined in a one-way ANOVA with three levels: 1) persons with average and above-average sleep duration who did not report insomnia symptoms (Non-Insomnia subgroup), 2) persons with short sleep duration who do not report insomnia symptoms (SSD subgroup), and 3) persons who report insomnia symptoms (INSD and ISS subgroups combined). In the case of a significant main effect of sleep phenotype, the hypothesized additive effect of insomnia on objective short sleep duration, as compared to normal subjective and objective sleep, was examined through a one-way ANOVA with three levels (non-insomnia, non-insomnia with normal sleep duration, insomnia short sleep duration).

The dependence of these effects upon various covariates was assessed in supplementary ANCOVA models. Omnibus tests were evaluated at p < 0.05, whereas follow up pairwise comparisons were assessed at Bonferroni-adjusted p = 0.05/2 = 0.025 (controlling for planned comparisons between the non-insomnia and each of the two other sleep subgroups in each ANOVA model).

The following covariates were entered in all models: 1) socio-demographic variables (gender, age, education level, living alone), 2) BMI, 3) diagnosis of depression, severity of self-reported depression (total score on the Geriatric Depression Scale; GDS) or anxiety symptoms (total score on the appropriate Hospital Anxiety and Depression scale; HADS), use of psychotropic medications, i.e., antipsychotics, antidepressants, and sleep inducing agents such as benzodiazepines, zolpidem, trazodone, and mirtazapine, multiple (≥4) physical comorbidities, 4) lack of physical activity, and 5) sleep apnea symptoms. All analyses were conducted using SPSS 25.0 (IBM Corp., Armonk, NY).

RESULTS

Association of MCI versus CNI with cortisol levels

Table 1 presents demographic and clinical characteristics by MCI versus CNI group. On average, MCI compared to CNI participants were older (p = 0.005), less educated (p = 0.001), reported more severe symptoms of depression (GDS; p = 0.004), more likely to be diagnosed with clinical depression (p = 0.001), and use psychotropic medication (40.3% versus 26.2%, p = 0.025). Regarding sleep characteristics, the MCI had longer average sleep onset latency in comparison to the CNI group (p = 0.018), and longer nighttime awakenings (p = 0.017). No significant differences were observed regarding sleep efficiency, total sleep time, wake time after sleep onset, and percentage of insomnia symptoms between the MCI and CNI subjects (all p > 0.178). The two groups were also comparable on gender distribution, self-reported anxiety symptomatology, BMI, percentage of persons living alone, following a sedentary lifestyle, and suffering from multiple comorbidities (p > 0.2).

Table 1

Demographic, clinical, and sleep characteristics by diagnostic group

	CNI (n = 84)	MCI (n = 124)	p
Age (y)	73.18±7.36	76.0±6.7	0.005
Female gender	52 (61.9%)	87 (70.2%)	0.232
Education (y)	5.85±2.89	4.47±3.06	0.001
BMI (kg/m²)	29.58±5.12	29.58±4.68	0.574
Lives alone	17 (20.2%)	33 (26.6%)	0.3
Lack of physical activity	30 (35.7%)	37 (29.8%)	0.3
Multiple comorbidities	20 (23.8%)	32 (25.8%)	0.8
MMSE	26.72±3.02	22.69±2.87	0.001
GDS	3.1±3.27	4.53±3.58	0.004
Depression	16 (19.0%)	50 (40.3%)	0.001
HADS	2.96±3.64	2.86±2.94	0.816
Psychotropic medication	22 (26.2%)	50 (40.3%)	0.039
Cortisol (mg/ml)	50.8±82.0	72.8±96.3	0.009
Sleep onset latency (min)	10.75±7.75	14.72±13.78	0.018
Sleep efficiency (%)	81.35±7.67	80.4±9.11	0.430
Total sleep time (min)	445.69±87.61	438.13±85.33	0.536
Wake after sleep onset (min)	84.31±43.5	83.81±45.41	0.937
Nighttime awakenings (min)	5.8±2.55	6.8±3.18	0.017
Insomnia symptoms (%)	32.1±47	36±48	0.548
Sleep apnea symptoms (%)	19.0%	22.3%	0.577

CNI, cognitive non-impairment; MCI, mild cognitive impairment; BMI, body mass index; MMSE, Mini-Mental State Examination; GDS, Geriatric Depression Scale; HADS, Hospital Anxiety and Depression Scale. Unless otherwise indicated, values represent mean±SD.

Importantly, average morning cortisol (log transformed) was higher in the MCI as compared to the CNI group (p = 0.009) and this difference remained significant after controlling for age, gender, BMI, education, depression diagnosis, and psychotropic medications (p = 0.013).

Sleep and clinical characteristics by insomnia phenotype group in participants with MCI

In this and the following section, we focus on the MCI group since in the CNI group no differences of cortisol levels were observed among the different sleep phenotypes. On the basis of sleep efficiency indices and self-reported sleep complaints, MCI participants were assigned to five groups: 1) Non-insomnia group (n = 42), 2) Short-sleep duration without insomnia group (SSD, n = 38), 3) Insomnia group (n = 44), 4) Insomnia-normal sleep duration group (INSD, n = 22), and 5) Insomnia-Short sleep duration group (ISS, n = 22). As shown in Table 2, the five groups were comparable on age, gender distribution, BMI, percentage of persons living alone, following a sedentary lifestyle, and suffering from multiple comorbidities (p≥0.05).

Table 2

Demographic, clinical, and sleep characteristics by insomnia phenotype among persons with mild cognitive impairment

	1. Non-insomnia n = 42	2. SSD n = 38	3. Insomnia n = 44	p		3a. INSD n = 22	3b. ISS n = 22	p
				1 versus 2	1 versus 3			1 versus 3a	1 versus 3b
Age (y)	74.62±7.23	76.21±6.25	77.14±6.47	0.3 (0.23)	0.09 (0.37)	76.45±6.64	77.82±6.38	0.3 (0.26)	0.07 (0.47)
Female gender	29 (69.0%)	22 (57.8%)	36 (81.8%)	0.3	0.2	18 (81.8%)	18 (81.8%)	0.5	0.5
Education (y)	4.74±3.05	5.39±3.43	3.41±2.44	0.4 (0.2)	0.03 (0.48)	2.95±2.34	3.86±2.49	0.02 (0.66)	0.2 (0.32)
BMI (kg/m²)	29.65±4.96	29.36±3.76	29.71±5.18	0.8 (0.07)	0.9 (0.01)	28.71±4.75	30.8±5.53	0.6 (0.19)	0.4 (0.22)
Lives alone	7 (16.7%)	8 (21.1%)	16 (38.1)	0.8	0.05	8 (36.4%)	8 (36.4%)	0.07	0.07
Lack of Physical Activity	12 (35.3%)	7 (36.8%)	13 (34.2%)	0.4	0.9	9 (45.0%)	4 (22.2%)	0.3	0.4
Multiple comorbidities	14 (33.3%)	7 (18.4%)	11 (25.0%)	0.1	0.5	4 (18.2%)	6 (27.3%)	0.3	0.8
MMSE	22.48±3.24	23.42±2.75	22.27±2.52	0.17 (0.3)	0.75 (0.07)	22.36±2.72	22.18±2.36	0.9 (0.04)	0.7 (0.11)
GDS	3.42±3.23	4.16±3.31	5.91±3.73	0.3 (0.23)	0.002 (0.71)	6.23±3.86	5.76±3.68	0.003 (0.79)	0.01 (0.68)
Depression	18 (42.9%)	9 (23.7%)	23 (52.3%)	0.07	0.4	13 (59.1%)	10 (45.5%)	0.2	0.8
HADS	2.22±2.63	2.19±2.16	4.02±3.04	0.9 (0.01)	0.005 (0.63)	4.47±3.2	3.71±2.92	0.002 (0.77)	0.04 (0.54)
Psychotropics	18 (42.6%)	11 (28.9%)	21 (47.7%)	0.2	0.7	13 (59.1%)	8 (36.4%)	0.2	0.6
Sleep onset latency (min)	8.85±8.32	19.76±15.44	15.96±14.65	< 0.001 (0.88)	0.007 (0.6)	11.56±10.43	20.35±17.04	0.4 (0.29)	0.001 (0.86)
Sleep efficiency (%)	87.22±2.9	73.56±6.4	79.77±10.31	< 0.001	< 0.001	88.17±4.2	71.37±7.3	0.5	< 0.001
Night total sleep time (min)	430.86±73.3	398.21±66.4	379.97±81.74	0.04 (0.47)	0.003 (0.66)	418.17±67.9	343.13±79.9	0.5 (0.18)	< 0.001 (1.14)
Wake after sleep onset (min)	54.5±21.2	122.03±40.2	78.78±43.83	< 0.001 (2.1)	0.002 (0.71)	45.61±21.2	111.96±34.6	0.3 (0.42)	< 0.001 (2)
Nighttime awakenings (min)	5.44±2.45	7.77±2.69	7.26±3.76	< 0.001 (0.91)	0.01 (0.57)	5.47±1.71	9.05±4.39	0.9 (0.01)	< 0.001 (1.02)
Sleep apnea symptoms (%)	6 (14.3%)	5 (13.2%)	14 (35%)	0.9	0.07	6 (27.3%)	8 (36.4%)	0.3	0.05
Cortisol (ng/ml)	46.1±54.9	89.7±112.1	90.02±114.7	0.045 (0.49)	0.014 (0.49)	50.5±50.3	115.9±137.9	0.4 (0.08)	0.011 (0.67)

BMI, body mass index; MMSE, Mini-Mental State Examination; GDS, Geriatric Depression Scale; HADS, Hospital Anxiety and Depression Scale; INSD, Insomnia-normal sleep duration; SSD, Short sleep duration (without insomnia); ISS, Insomnia with short sleep duration. p-values reported correspond to independent samples t-tests between the sub-groups presented, with the corresponding effect sizes (Cohen’s d) in the parentheses.

Regarding actigraphy indices, participants in the SSD and ISS groups displayed shorter nighttime TST, higher sleep latency, and wake after sleep onset than the non-insomnia group (in addition to lower night sleep efficiency). The SSD and ISS groups were comparable on all sleep parameters (p > 0.1) with the exception of nighttime TST (SSD > ISS, p = 0.005). Moreover, participants in the ISS group reported more symptoms of anxiety than SSD participants (p = 0.04) and were more likely to carry a diagnosis of depression (45.5 versus 23.7%; p = 0.08). Although, as expected, the non-insomnia and INSD groups did not differ on objective sleep parameters (p > 0.3), participants who reported insomnia symptoms also reported more depressive-based on GDS-symptoms (p = 0.003) and anxiety-based on HADS- symptoms (p = 0.002) without, however, being significantly more likely to have a diagnosis of depression (p = 0.2).

Association of insomnia symptoms and short sleep duration with cortisol levels in participants with MCI

In the first step of the analysis, we assessed differences on log cortisol levels due to short sleep duration (SSD subgroup) and insomnia (INSD and ISS subgroups combined) on log cortisol levels revealing a significant main effect of sleep phenotypes [F(2,119) = 3.62, p = 0.030]. Pairwise comparisons revealed significantly elevated cortisol levels among MCI patients with insomnia as compared to the non-insomnia subgroup (p = 0.014) (Table 2), whereas cortisol levels were only marginally elevated among MCI persons with SSD as compared to persons who did not report insomnia symptoms and did not show objective short sleep duration (non-insomnia subgroup; p = 0.045) (Table 2). As shown in Fig. 2, log cortisol levels increase linearly across subgroups (linear trend: B = 0.232, p = 0.015).

Fig. 2

Log-transformed cortisol levels by sleep phenotype among persons with MCI, adjusted for age, education level, gender, depression diagnosis, and psychotropic medication. Non-insomnia, persons who do not report insomnia symptoms with normal sleep duration; SSD, persons without insomnia with actigraphy-based short sleep duration; INSD, persons with insomnia with actigraphy-based normal sleep duration; ISS, persons with insomnia with actigraphy-based short sleep duration. Brackets indicate significant differences at Bonferroni-adjusted p < 0.025. Vertical bars represent standard error.

Association of short sleep duration with cortisol levels in participants with insomnia symptoms and MCI

A follow up one-way ANOVA further supported the notion that short sleep duration and insomnia exert additive effects on cortisol levels. Thus the main effect of sleep phenotype was, again, significant [F(2,81) = 3.42, p = 0.038] with log cortisol levels increasing linearly across non-insomnia, INSD, and ISS subgroups (linear trend: B = 0.307, p = 0.011). Pairwise comparisons revealed that, as compared to the non-insomnia subgroup, cortisol levels were significantly elevated only in persons with the ISS phenotype (p = 0.011), whereas presence of insomnia with normal sleep duration was not associated with increased cortisol levels (p = 0.2) (Table 2). Both linear trends and all pairwise comparisons specified above remained significant at p = 0.01 after controlling for the following sets of variables: 1) socio-demographics and lifestyle variables (age, education, gender, living alone, lack of physical activity, BMI), 2) depression diagnosis and psychotropic medication use, 3) HADS and GDS score, and 4) multiple comorbidities and sleep apneasymptoms.

We reclassified participants by defining phenotypes according to night TST and subjective insomnia symptoms. The overall group effect in the ANOVA is not significant (p = 0.1), but the linear trend is (p = 0.04). Pairwise comparisons indicated that log cortisol was higher in the ISS as compared to the non-insomnia (p = 0.04), and INSD (p = 0.04) groups. In the CNI group, p values for pairwise comparisons did not exceed p = 0.15 (linear trend: p > 0.8). These results are very similar when we used sleep efficiency to define phenotypes. Also the cut off for short sleep in this population based on actigraphy is very similar between the two methods of phenotyping, i.e., approximately 7 h.

As already stated, there were no main effects of sleep subtype on cortisol levels among CNI participants (p > 0.21 across tests).

Associations between cortisol levels and cognitive capacity

We explored associations between log cortisol levels and indices of cognitive capacity among MCI participants (total MMSE score, and three composite indices of episodic memory, attention/executive function, and language). Each composite index represented average age- and education-adjusted z scores across several neuropsychological tests based on Greek population norms. Results revealed a curvilinear association of cortisol with the memory composite (which included delayed as well as immediate verbal and visuospatial memory scores) among MCI participants comprising the short sleep phenotypes (combined ISS and SSD subgroups; R² = 0.087, p = 0.1) (Supplementary Figure 1). This trend was largely attributed to a significant curvilinear association of cortisol levels with the retention index of the Auditory Verbal Learning Test (R² = 0.147, p = 0.022). Associations with other cognitive indices and MMSE were in the weak range (R² < 0.06) (Supplementary Figures 1 and 2).

DISCUSSION

The primary finding of this cross-sectional study is that basal plasma levels of cortisol in participants with MCI are elevated as compared to cognitively non-impaired controls and this increase is primarily associated with the SSD and ISS phenotypes.

Recent reviews and meta-analyses concluded that CSF but not plasma or saliva cortisol levels are moderately elevated in patients with MCI [26], whereas plasma cortisol levels were reported to be elevated only in patients with MCI due to Alzheimer’s disease, a more severe type of pre-dementia MCI [27]. The inconsistency between our findings and previous studies regarding plasma levels of cortisol in patients with MCI is primarily related to the fact that most of these studies included a small sample size, heterogeneous samples, different time of plasma sampling, and age differences [26]. Furthermore, to our knowledge, our study is the first one to demonstrate that this elevation is strongly and significantly associated with the ISS and in a lesser degree to SSD phenotype. It appears that the combination of the subjective complaint of insomnia, an index of emotional stress, with an objective measure of short sleep duration has a synergistic effect on cortisol levels. These results are consistent with our previously reported findings that only the ISS phenotype and not the insomnia normal sleep duration (INS) phenotype is associated with activation of the HPA axis [50, 51].

Based on our results short sleep duration without the subjective complaint of insomnia (SSD) is also associated with higher basal cortisol levels. Many previous studies based primarily on self-reported measures have indicated that short sleep duration is associated with adverse effects on cardiovascular function [30 –33] and memory [34, 35]. In our study, we failed to demonstrate that the SSD and ISS phenotypes are associated with higher basal cortisol levels in the CNI group. This may be related to the fact that the sample of CNI and particularly the ISS phenotype was small, and the cortisol values recorded had high variance. Another possibility is that in patients with MCI the association of cortisol with the ISS and SSD phenotypes is stronger compared to cognitively non-impaired subjects due to a higher degree of HPA axis dysregulation, i.e., impaired negative glucocorticoid feedback mechanism in this group. Future studies with larger samples and more frequent sampling of cortisol levels particularly in the evening when the HPA axis is quiescent are needed to test the above hypothesis.

We have previously demonstrated that the ISS phenotype is associated with neuropsychological impairment in a general population sample [52]. Similar results have been reported by other investigators defining ISS phenotype on either objective or self-reported measures [53, 54]. In this study ISS phenotype in patients with MCI is associated with high cortisol levels which have been shown to be negatively affecting memory and executive functions. Exploration of associations between log cortisol levels and indices of cognitive capacity among MCI participants in this study revealed a curvilinear association of cortisol with the memory composite (which included delayed as well as immediate verbal and visuospatial memory scores) among MCI participants comprising the short sleep phenotypes. We speculate that cortisol is the mediating factor of the negative impact of the ISS and SSD phenotypes on memory function in patients with MCI.

Previous literature including our group reports that MCI is quite prevalent in the general population of persons 60 years or older [3 –6] and about 10–15% per year tend to convert to dementia [3 –5]. Based on our findings, we hypothesize that it is possible that the SSD and ISS phenotypes which are present in about 50% of patients with MCI may be major risk factors leading to progression of MCI to dementia.

Studies in the past including ours have reported an inverse U-curve association between sleep duration and cognition in cognitively impaired elderly [55]. Also, we have shown that in patients with MCI long sleep duration is an indicator of worse cognitive performance [45]. That both extremes of sleep duration are associated with cognitive impairment can be interpreted by that SSD and ISS phenotypes via increased basal cortisol levels are risk factors leading to cognitive impairment, whereas prolonged sleep may be a marker of worse cognitive performance among patients with MCI.

In our study within MCI participants, depression, a condition that is associated in its severe form with activation of the HPA axis, was more frequent in the normal sleep duration phenotype compared to the ISS phenotype. This suggests that the effect of the ISS phenotype on the cortisol levels does not relate to a mental disorder such as depression.

Our findings have important clinical implications given that high cortisol may be a risk factor of progression of MCI to dementia. It has been reported that antidepressant medications, such as trazodone, or doxepin when used, in low dosages as sleep inducing agents, can lower the peripheral levels of cortisol in patients with the ISS insomnia phenotype [56, 57]. Interestingly, it has been reported in a retrospective study from the University of California San Francisco that the use of trazodone in low dosage in patients with cognitive impairment and sleep complaints slowed down the progress of cognitive decline in a follow up period of four years [58]. However, another study based on UK population-based electronic health records failed to find association between reduced risk of dementia and trazodone use in comparison to other antidepressants [59]. Also, other medications such as benzodiazepine receptor agonists have not shown to have an effect on cortisol levels [60 –62]. Also, cognitive behavioral therapy for insomnia does not appear to have an effect in the cortisol levels in patients suffering from the ISS phenotype [56]. Future studies should examine the effect of low doses of these antidepressants on the sleep duration, cortisol levels, and cognitive decline in patients with MCI and ISS insomnia phenotype.

This study has several strengths, including a relatively large sample for the depth of information acquired per participant, detailed neuropsychological and neuropsychiatric assessment, and objective sleep measures. In addition, the sample of this study is a naturalistic one, representing community-dwelling MCI patients without exclusion criteria that may result in a select group of patients with cognitive impairment. Our study also has some limitations. First, sleep recording was based on actigraphy and not polysomnography (PSG) which is the gold-standard for sleep efficiency and structure and for detection of other sleep disorders, such as sleep apnea which is highly prevalent in elderly and is a potential confounding factor in cortisol levels [63]. Given the limitation of actigraphy to differentiate between sleep and sedentary time in bed, sleep duration in this group may have been overestimated. Second, actigraphy was conducted for only 3 nights, while in other studies 1 to 2 weeks of actigraphy recordings were used. Third, basal cortisol levels were assessed using a single morning blood sample and were not based in individual 24 h diurnal profiles of cortisol, a method more suitable for small intense experimental physiological studies. Finally, in this study, the cut-off sleep duration that separates the ISS phenotype from the INS phenotype, using either median sleep efficiency or TST, is about 7 h. Although this cut off is based on a solid mathematical approach used in all our previous studies to distinguish short from normal sleepers, is different than the cut-off of 6 h, which was the median PSG sleep duration in physiological studies and in large random general population samples [42 , 51]. It is to be expected that the median value of objective sleep duration will differ based on the method used (i.e., actigraphy versus PSG), population studied (i.e., general random sample versus clinical or volunteer), and age of the sample [56].

Conclusion

In conclusion, to our knowledge this is the first case-control, cross-sectional study to demonstrate that peripheral basal cortisol levels are elevated in persons with MCI and this elevation is primarily associated by the ISS and SSD phenotypes. It appears that these two phenotypes are major risk factors for the progression of MCI to dementia. Future clinical trials should assess whether medications that appear to lengthen sleep duration and lower cortisol levels, e.g., trazodone, may be effective in reversing/delaying the progression of MCI into dementia.

Footnotes

ACKNOWLEDGMENTS

We thank study coordinator Cynthia Manasaki for her continuing support. We also thank all Primary Care Physicians from the Primary Health Care facilities who participated in this study. The Cretan Aging Cohort was instituted with funding from the National Strategic Reference Framework (ESPA) 2007–2013, Program: THALES, University of Crete, entitled: “A multi-disciplinary network for the study of Alzheimer’s Disease” (Grant: MIS 377299) to AV. Also. D.A. was supported by the co-financed from Greece and European Union (European Social Fund -ESF) operational program “Human Resources, Development, Education and Lifelong Learning”, with the action “Support to Post-doctoral Researchers -Cycle B” (MIS 5033021), implemented by IK

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Merlino

, Piani

, Gigli

, Cancelli

, Rinaldi

, Baroselli

, Serafini

, Zanchettin

, Valente

(2010) Daytime sleepiness is associated with dementia and cognitive decline in older Italian adults: A population-based study. Sleep Med 11, 372–377.

https://www.nationmaster.com/country-info/profiles

Roberts

, Knopman

(2013) Classification and epidemiology of MCI. Clin Geriart Med 29, 753–72.

Petersen

(2011) Mild cognitive impairment. N Engl J Med 364, 2227–2234.

Petersen

(2016) Mild cognitive impairment. Continuum (Minneap Minn) 22, 404–418.

Zaganas

, Simos

, Basta

, Kapetanaki

, Panagiotakis

, Koutentaki

, Fountoulakis

, Bertsias

, Duijker

, Tziraki

, Scarmeas

, Plaitakis

, Boumpas

, Lionis

, Vgontzas

(2019) The Cretan Aging Cohort: Cohort description and burden of dementia and mild cognitive impairment. Am J Alzheimers Dis Other Demen 34, 23–33.

Yannakoulia

, Kontogianni

, Scarmeas

(2015) Cognitive health and Mediterranean diet: Just diet or lifestyle pattern? Ageing Res Rev 20, 74–78.

Lupien

, de Leon

, de Santi

, Convit

, Tarshish

, Nair

, Thakur

, McEwen

, Hauger

, Meaney

(1998) Cortisol levels during human aging predict hippocampal atrophy and memory deficits. Nat Neurosci 1, 69–73.

Sapolsky

, Krey

, McEwen

(1986) The neuroendocrinology of stress and aging: The glucocorticoid cascade hypothesis. Endocr Rev 7, 284–301.

10.

Vgontzas

, Zoumakis

, Bixler

, Lin

, Prolo

, Vela-Bueno

, Kales

, Chrousos

(2003) Impaired nighttime sleep in healthy old versus young adults is associated with elevated plasma interleukin-6 and cortisol levels: Physiologic and therapeutic implications. J Clin Endocrinol Metab 88, 2087–2095.

11.

McEwen

(2002) Sex, stress and the hippocampus: Allostasis, allostatic load and the aging process. Neurobiol Aging 23, 921–939.

12.

Chrousos

(2009) Stress and disorders of the stress system. Nat Rev Endocrinol 5, 374–381.

13.

Ross

, Jones

, Blockman

(2018) We are tired of ‘adrenal fatigue’. S Afr Med J 108, 724–725.

14.

Lupien

, McEwen

, Gunnar

, Heim

(2009) Effects of stress throughout the lifespan on the brain, behaviour and cognition. Nat Rev Neurosci 10, 434–445.

15.

Huang

, Lui

, Chang

, Lu

, Wang

, Chang

(2009) Elevated basal cortisol level predicts lower hippocampal volume and cognitive decline in Alzheimer’s disease. J Clin Neurosci 16, 1283–1286.

16.

Arnsten

(2009) Stress signaling pathways that impair prefrontal cortex structure and function. Nat Rev Neurosci 10, 410–422.

17.

Stomby

, Boraxbekk

, Lundquist

, Nordin

, Nilsson

, Adolfsson

, Nyberg

, Olsson

(2016) Higher diurnal salivary cortisol levels are related to smaller prefrontal cortex surface area in elderly men and women. Eur J Endocrinol 175, 117–126.

18.

Lee

, Glass

, McAtee

, Wand

, Bandeen-Roche

, Bolla

, Schwartz

(2007) Associations of salivary cortisol with cognitive function in the Baltimore memory study. Arch Gen Psychiatry 64, 810–818.

19.

Comijs

, Gerritsen

, Penninx

, Bremmer

, Deeg

, Geerlings

(2010) The association between serum cortisol and cognitive decline in older persons. Am J Geriatr Psychiatry 18, 42–50.

20.

Geerlings

, Sigurdsson

, Eiriksdottir

, Garcia

, Harris

, Gudnason

, Launer

(2015) Salivary cortisol, brain volumes, and cognition in community-dwelling elderly without dementia. Neurology 85, 976–983.

21.

Popp

, Schaper

, Kölsch

, Cvetanovska

, Rommel

, Klingmüller

, Dodel

, Wüllner

, Jessen

(2009) CSF cortisol in Alzheimer’s disease and mild cognitive impairment. Neurobiol Aging 30, 498–500.

22.

Popp

, Wolfsgruber

, Heuser

, Peters

, Hüll

, Schröder

, Möller

, Lewczuk

, Schneider

, Jahn

, Luckhaus

, Perneczky

, Frölich

, Wagner

, Maier

, Wiltfang

, Kornhuber

, Jessen

(2015) Cerebrospinal fluid cortisol and clinical disease progression in MCI and dementia of Alzheimer’s type. Neurobiol Aging 36, 601–607.

23.

Wolf

, Convit

, Thorn

, de Leon

(2002) Salivary cortisol day profiles in elderly with mild cognitive impairment. Psychoneuroendocrinology 27, 777–789.

24.

Lind

, Edman

, Nordlund

, Olsson

, Wallin

(2007) Increased saliva cortisol awakening response in patients with mild cognitive impairment. Dement Geriatr Cogn Disord 24, 389–395.

25.

Arsenault-Lapierre

, Chertkow

, Lupien

(2010) Seasonal effects on cortisol secretion in normal aging, mild cognitive impairment and Alzheimer’s disease. Neurobiol Aging 31, 1051–1054.

26.

Zheng

, Tal

, Yang

, Middleton

, Udeh-Momoh

(2020) Cortisol hypersecretion and the risk of Alzheimer’s disease: A systematic review and meta-analysis. Ageing Res Rev 64, 101171.

27.

Ouanes

, Popp

(2019) High cortisol and the risk of dementia and Alzheimer’s disease: A review of the literature. Front Aging Neurosci 11, 43.

28.

Neikrug

, Ancoli-Israel

(2010) Sleep disorders in the older adult- a mini-review. Gerontology 56, 181–189.

29.

Vitiello

, Prinz

(19890 Alzheimer’s disease. Sleep and sleep/wake patterns. Clin Geriatr Med 5, 289–299.

30.

Vgontzas

, Liao

, Bixler

, Chrousos

, Vela-Bueno

(2009) Insomnia with objective short sleep duration is associated with a high risk for hypertension. Sleep 32, 491–497.

31.

Kundel

, Reid

, Fayad

, Ayappa

, Mani

, Rueschman

, Redline

, Shea

, Shah

(2021) Sleep duration and vascular inflammation using hybrid positron emission tomography/magnetic resonance imaging: Results from the Multi-Ethnic Study of Atherosclerosis (MESA). J Clin Sleep Med 17, 2009–2018.

32.

Abdalla

, Schwartz

, Cornelius

, Chang

, Alcántara

, Shechter

(2020) Objective short sleep duration and 24-hour blood pressure. Int J Cardiol Hypertens 7, 100062.

33.

Domínguez

, Fuster

, Fernández-Alvira

, Fernández-Friera

, López-Melgar

, Blanco-Rojo

, Fernández-Ortiz

, García-Pavía

, Sanz

, Mendiguren

, Ibañez

, Bueno

, Lara-Pezzi

, Ordovás

(2019) Association of sleep duration and quality with subclinical atherosclerosis. J Am Coll Cardiol 73, 134–144.

34.

Miyata

, Noda

, Iwamoto

, Kawano

, Okuda

, Ozaki

(2013) Poor sleep quality impairs cognitive performance in older adults. J Sleep Res 22, 535–541.

35.

McSorley

, Bin

, Lauderdale

(2019) Associations of sleep characteristics with cognitive function and decline among older adults. Am J Epidemiol 188, 1066–1075.

36.

Basta

, Simos

, Bertsias

, Duijker

, Zaganas

, Koutentaki

, Anastasaki

, Mavroidis

, Kalomoiri

, Panagiotakis

, Lionis

, Vgontzas

(2018) Association between insomnia symptoms and cognitive impairment in the Cretan Aging Cohort. Eur Geriatr Med 9, 697–706.

37.

Jelicic

, Bosma

, Ponds

RWH

, Van Boxtel

, Houx

, Jolles

(2002) Subjective sleep problems in later life as predictors of cognitive decline. Int J Geriatr Psychiatry 17, 73–77.

38.

Cricco

, Simonisick

, Foley

(2001) The impact of insomnia on cognitive functioning in older adults. J Am Geriatr Soc 49, 1185–1189.

39.

Shi

, Chen

, Ma

, Bao

, Han

, Wang

, Shi

, Vitiello

, Lu

(2018) Sleep disturbances increase the risk of dementia: A systematic review and meta-analysis. Sleep Med Rev 40, 4–16.

40.

Leng

, McEvoy

, Allen

, Yaffe

(2017) Association of sleep-disordered breathing with cognitive function and risk of cognitive impairment: A systematic review and meta-analysis. JAMA Neurol 74, 1237–1245.

41.

Fernandez-Mendoza

, Vgontzas

, Liao

, Shaffer

, Vela-Bueno

, Basta

, Bixler

(2012) Insomnia with objective short sleep duration and incident hypertension: The Penn State cohort. Hypertension 60, 929–935.

42.

Vgontzas

, Liao

, Pejovic

, Calhoun

, Karataraki

, Bixler

(2009) Insomnia with objective short sleep duration is associated with type 2 diabetes: A population based study. Diabetes Care 32, 1980–1985.

43.

Vgontzas

, Liao

, Pejovic

, Calhoun

, Karataraki

, Basta

, Fernández-Mendoza

, Bixler

(2010) Insomnia with short sleep duration and mortality: The Penn State cohort. Sleep 33, 1159–1164.

44.

Vgontzas

, Fernandez-Mendoza

, Liao

, Bixler

(2013) Insomnia with objective short sleep duration: The most biologically severe phenotype of the disorder. Sleep Med Rev 17, 241–254.

45.

Basta

, Simos

, Vgontzas

, Koutentaki

, Tziraki

, Zaganas

, Panagiotakis

, Kapetanaki

, Fountoulakis

, Lionis

(2019) Associations between sleep duration and cognitive impairment in MCI. J Sleep Res 28, 12864.

46.

Fountoulakis

, Tsolaki

, Chantzi

, Kazis

(2000) Mini Mental State Examination (MMSE): A validation study in Greece. Am J Alzheimers Dis Other Demen 15, 342–345.

47.

Dardiotis

, Kosmidis

, Yannakoulia

, Hadjigeorgiou

, Scarmeas

(2014) The Hellenic Longitudinal Investigation of Aging and Diet (HELIAD): Rationale, study design, and cohort description. Neuroepidemiology 43, 9–14.

48.

Winblad

, Palmer

, Kivipelto

, Jelic

, Fratiglioni

, Wahlund

, Nordberg

, Bäckman

, Albert

, Almkvist

, Arai

, Basun

, Blennow

, de Leon

, DeCarli

, Erkinjuntti

, Giacobini

, Graff

, Hardy

, Jack

, Jorm

, Ritchie

, van Duijn

, Visser

, Petersen

(2004) Mild cognitive impairment-beyond controversies, towards a consensus: Report of the International Working Group on Mild Cognitive Impairment. J Intern Med 256, 240–246.

49.

Bixler

, Vgontzas

, Lin

, Vela-Bueno

, Kales

(2002) Insomnia in Central Pennsylvania. J Psychosom Res 53, 589–592.

50.

Vgontzas

, Tsigos

, Bixler

, Stratakis

, Zachman

, Kales

, Vela-Bueno

, Chrousos

(1998) Chronic insomnia and activity of the stress system: A preliminary study. J Psychosom Res 45, 21–31.

51.

Vgontzas

, Bixler

, Lin

, Prolo

, Mastorakos

, Vela-Bueno

, Kales

, Chrousos

(2001) Chronic insomnia is associated with nyctohemeral activation of the hypothalamic-pituitary-adrenal axis: Clinical implications. J Clin Endocrinol Metab 86, 3787–3794.

52.

Fernandez-Mendoza

, Vgontzas

(2013) Insomnia and its impact on physical and mental health. Curr Psychiatry Rep 15, 418.

53.

Baril

, Beiser

, Sanchez

, Mysliwiec

, Redline

, Gottlieb

, O’Connor

, Gonzales

, Himali

, Seshadri

, Himali

, Pase

(2022) Insomnia symptom severity and cognitive performance: Moderating role of APOE genotype. Alzheimers Dement 18, 408–421.

54.

Olaithe

, Ree

, McArdle

, Donaldson

, Pushpanathan

, Eastwood

, Bucks

(2021) Cognitive dysfunction in insomnia phenotypes: Further evidence for different disorders. Front Psychiatry 12, 688672.

55.

Basta

, Koutentaki

, Vgontzas

, Zaganas

, Vogiatzi

, Gouna

, Bourbouli

, Panagiotakis

, Kapetanaki

, Fernandez-Mendoza

, Simos

(2020) Objective daytime napping is associated with disease severity and inflammation in patients with mild to moderate dementia. J Alzheimers Dis 74, 803–815.

56.

Vgontzas

, Puzino

, Fernandez-Mendoza

, Krishnamurthy

, Basta

, Bixler

(2020) Effects of trazodone versus cognitive behavioral therapy in the insomnia with short sleep duration phenotype: A preliminary study. J Clin Sleep Med 16, 2009–2019.

57.

Rodenbeck

, Cohrs

, Jordan

, Huether

, Rüther

, Hajak

(2003) The sleep-improving effects of doxepin are paralleled by a normalized plasma cortisol secretion in primary insomnia. A placebo-controlled, double-blind, randomized, cross-over study followed by an open treatment over 3 weeks. Psychopharmacology (Berl) 170, 423–428.

58.

, Walsh

, Neylan

, Vossel

, Yaffe

, Krystal

, Miller

, Karageorgiou

(2019) Long-term trazodone use and cognition: A potential therapeutic role for slow-wave sleep enhancers. J Alzheimers Dis 67, 911–921.

59.

Brauer

, Lau

WCY

, Hayes

, Man

KKC

, Osborn

DPJ

, Howard

, Kim

, Wong

ICK

(2019) Trazodone use and risk of dementia: A population-based cohort study. PLoS Med 16, e1002728.

60.

Mikkelsen

, Søderman

, Kiss

, Mirza

(2005) Effects of benzodiazepines receptor agonists on the hypothalamic-pituitary-adrenocortical axis. Eur J Pharmacol 519, 223–230.

61.

Mikkelsen

, Bundzikova

, Larsen

, Hansen

, Kiss

(2008) GABA regulates the rat hypothalamic-pituitary-adrenocortical axis via different GABA-A receptor alpha-subtypes. Ann N Y Acad Sci 1148, 384–392.

62.

Pechnick

, Lacayo

, Manalo

, Bholat

, Spivak

(2011) Eszopiclone stimulates the hypothalamo-pituitary-adrenal axis in the rat. Eur J Pharmacol 661, 22–26.

63.

Bixler

, Vgontzas

, Ten Have

, Tyson

, Kales

(1998) Effects of age on sleep apnea in men: I. Prevalence and severity. Am J Respir Crit Care Med 157, 144–148.