Sleep is an active phenomenon regulated by a highly integrated network of cortical and subcortical structures. The efficiency of this complex pattern may be compromised at various levels during physiological aging [1]. About half of elderly subjects report disturbed sleep, and there is extensive evidence that sleep alterations are closely linked to neurodegenerative disorders [2, 3]. Sleep and wake disturbances are common among people with dementia. Up to 70% of patients with early-stage dementia report sleep disturbances [4]. Among patients with Alzheimer’s disease (AD), disturbed sleep is associated with poorer daily functioning, aggression and agitation. In general, sleep alterations seem to begin in the early stages of cognitive impairment and tend to worsen as the disease progresses [5, 6].

The association between sleep and dementia is complex and probably bidirectional [4, 7]. Sleep is involved in maintaining the anatomical integrity of the brain through different and complex mechanisms. Sleep problems may contribute to neurodegenerative changes. Consequently, the preservation of restorative sleep is strategic for memory consolidation through the transfer of information from the hippocampus to the anterior regions of the brain [7]. In addition, the role of sleep in synaptic plasticity and the promotion of amyloid (AB) removal has been demonstrated [4, 8].

Neurodegenerative processes can lead to disturbed sleep, and the coexistence of sleep disorders and dementia has been associated with a more rapid decline in cognitive performance [8]. Therefore, the demonstration that potentially correctable sleep disturbances can be recognized at an early stage of cognitive decline would be of interest because of possible implications in developing treatment strategies [9].

Actigraphic assessment allows a simple determination of objective sleep changes. For example, it has been shown that in approximately half of the general population of older adults, sleep structure tends to become progressively disorganized [8], and actigraphic studies in community-dwelling people with AD have shown a deterioration of rest-activity cycles [10].

In this preliminary research, we evaluated actigraphic patterns in three populations of subjects with a distinct state of cognitive efficiency to test whether the characterization of sleep parameters can be specifically associated with the extent of cognitive decline.

Of 59 patients initially recruited from among those with mild cognitive impairment, 36 were excluded due to the presence of one or more exclusion criteria while three refused to participate or discontinued the actigraphic assessment. In particular, 5 patients with MCI and 4 with AD were excluded due to the presence of sleep disturbances. Twenty patients, 10 with MCI and 10 with AD were included together with ten controls. We had no missing values for any of the measurements taken. All included subjects were able to complete the 7-day actigraphic study without any extension of the recording. The results obtained at PSQI with the detailed description of the 7 components are reported in Table 1. The power analysis resulted in an 80% probability of rejecting the H0 hypothesis when it was false. When analysing age, PSQI, TIB, TST, SL, SE, WASO, NA and MA with the Kolmogorov-Smirnov test in three groups of patients, we observed that this test was non-significant in all three groups for each variable analysed, suggesting a normal distribution. The three groups resulted similar for baseline characteristics, as sex (chi-squared = 0.000; df = 2; p = 1.000), hypertension (chi-squared = 0.800; df = 2; p = 0.670), diabetes (chi-squared = 0.373; df = 2; p = 0.830), dyslipidaemia (chi-squared = 0.287; df = 2; p = 0.866), smoke (chi-squared = 0.341; df = 2; p = 0.843) and BMI (chi-squared = 0.952; df = 2; p = 0.621), as shown in Table 2. MCI and AD patients did not significantly differ for the prevalence of APOE $ϵ$4 allele (chi-squared = 1.818; df = 1; p = 0.178). ANOVA showed a significant difference between CS and MCI in SL (F = 13.095; p = 0.0001). CS and AD resulted significantly different in PSQI score (F = 4.461; p = 0.021), TIB (F = 6.113; p = 0.006), SE (F = 20.028; p = 0.0001), WASO (F = 10.030; p = 0.001), NA (F = 20.111; p = 0.0001), MA (F = 15.681; p = 0.0001), while TST did not reach statistical significance (F = 2.627; p = 0.091). MCI patients differed significantly from AD patients in SL, SE, WASO, NA and MA.

The GLM/Multivariate model resulted significantly, and we were able to reject the H${}_0$ hypothesis that the model explained zero variance in the dependent variables. The leading independent grouping variable resulted significantly both at multivariate test (p 0.0001). In contrast, the included covariates (age, sex, hypertension, smoking status, diabetes, dyslipidemia and obesity) did not significantly associate with the dependent variables at the multivariate test in this model. Estimated marginal means derived from the multivariate model are synthesized in Table 3, while the comparison among groups in the multivariate model is shown in Table 4.

Differences between MCI patients and CS were significant for SL (p = 0.05); AD patients had significantly worse score for PSQI (p = 0.01), TIB (p = 0.001), TST (p = 0.04), SL, SE, WASO, e MA (p = 0.0001) when compared to CS. When comparing AD and MCI, the differences was significant for SL (p = 0.01), WAS0 (p = 0.004), SE, NA and MA (p = 0.0001).

The results show that an actigraphic assessment can allow rapid and non-invasive detection of sleep changes in subjects with impaired cognitive performance. According to our results, reduced sleep quality may already be evident in patients with MCI. In agreement with the results of previous studies, the presence and extent of different sleep changes have become more frequent and severe in patients with AD [30, 31]. In particular, we found a progressive negative evolution of changes in SL and SE from CS to MCI to AD patients. Moreover, AD patients showed a higher prevalence of other alterations in actigraphic parameters, thus confirming that the evolution of cognitive impairment is associated with a reduction in sleep quality. The increase in the number of NA in cognitive impairment patients, besides underlining the relevance of actigraphy in detecting and defining the extent of motor hyperactivity [32], further emphasizes that sleep deprivation is a common feature in patients with dementia [33, 34, 35].

Some sleep alterations have been described in normal aging and reflect changes in sleep regulation processes [9]. Neurodegeneration may include neurons involved in sleep regulation. The alternation between sleep and wakefulness is regulated by a homeostatic component that considers the need for sleep in proportion to the duration of the waking state. On the other hand, the circadian process organizes the temporal distribution of wakefulness and sleep [9]. In the present study, we found a gradual increase in SL from controls to AD patients. This may suggest the occurrence of progressive alterations in the circadian process.

Sleep architecture is altered in the presence of sleep disorders. Therefore, in the present work, we excluded subjects with suspected primary sleep disorders [36].

The design of our study does not allow us to determine whether sleep alterations can be considered secondary or whether they may be involved in influencing the risk of cognitive impairment in patients. The interaction between sleep alterations and neurodegeneration is complex and bidirectional [4, 9]. The presence of sleep disorders in subjects with cognitive impairment, regardless of their interpretation as primary or secondary conditions, deserves attention for the pathophysiological implications. A correct approach and early diagnosis of sleep alterations are relevant to managing subjects at risk of developing dementia. In this regard, an interesting aspect that emerges from our study concerns the fact that the administration of the PSQI, a validated scale to assess sleep quality in the last month, was not able to detect significant differences between CS and MCI. Due to the low sample size, we did not perform a specific analysis to assess whether there were any differences in the different components of the PSQI between the two groups of patients. The different results between PSQI and actigraphy may be due to several reasons. First, the two approaches allow the acquisition of different data. Actigraphy information is based on objective data provided over 7 days, whereas PSQI supports subjective data over 1 month.

The advantage of providing objective data in a clinical setting where cognitive impairment may reduce the reliability of information based on patients’ ability to offer full cooperation is significant. Our data support the need to consider a multimodal approach in investigating sleep disorders in patients with cognitive impairment. Actigraphy seems to be a promising diagnostic approach for a fast, simple and widely available possibility to investigate and objectively detect sleep disorders, especially in the early stage of cognitive impairment. In agreement with our findings, the results of a recent survey of patients referred to a memory clinic showed that objective data supported by actigraphy are more reliable than self-reported information for assessing correlations between sleep disturbances and cognitive dysfunction [37]. Polysomnography remains the gold standard approach for definitive and accurate characterization of the sleep profile. Wider use of polysomnography is, however, limited for several reasons, including reduced patient compliance. In this regard, rapid screening with a more straightforward approach may be advantageous. Actigraphy, although mainly dedicated to the assessment of circadian rhythm alterations, providing reliable information on aspects related to the amount of sleep, could support the indication of the possibility of correcting parameters potentially implicated in the progression of cognitive impairment [10].

Our work has some limitations. The small sample size did not allow us to consider the possible influence of sex and age on sleep alteration even though the different groups of subjects in the study were comparable for these variables. Furthermore, the careful selection of subjects, mainly the exclusion of patients with sleep disorders, does not allow us to generalize our results to the entire population of patients with cognitive impairment. However, we wanted to obtain some preliminary indications on the possibility of extending the use of actigraphic assessments to obtain information on the association between sleep alterations and cognitive impairment. For this reason, we tried to select a relatively homogeneous group of patients. The fact that the subjects were not studied simultaneously must be considered to define our results as preliminary and capable of raising hypotheses and suggesting the need for further investigation into possible practical implications. For instance, the possibility that early correction of sleep disturbances in patients with MCI may reduce the risk of dementia conversion should be carefully considered when planning specifically designated investigations. It would be essential to assess sleep over 24 hours. We only evaluated sleep during the night without considering the 24-hour TST and the occurrence of naps, affecting the quality of the patients’ nightly sleep. In this regard, it has been reported that AD patients may have sleep episodes during the day [38]. Therefore, future studies should be performed that do not focus exclusively on changes in nocturnal sleep architecture. The delivery of information via PSQI, a self-administered questionnaire was obtained, in the case of AD patients, with the collaboration of caregivers responsible for assisting patients to correctly interpret each question. This approach can be considered at the potential risk of introducing inaccuracies. However, we have paid particular attention to minimize this possible bias by carefully monitoring the reliability of the caregivers and explaining in detail the meaning of the questions. Because subjective and objective measures can affect different aspects of sleep quality, each providing valuable insights, we recommend that both approaches, when examining sleep quality in older adults, be considered even in the presence of cognitive impairment [39]. The presence of mood disorders may significantly influence sleep characteristics. Their possible interference was not assessed in our study. Although the presence of a history of psychiatric illness was among the exclusion criteria, we cannot exclude the possibility that some sleep alterations observed in cognitively impaired patients may be related to a higher prevalence of anxiety and/or depression. Finally, all AD patients were on centrally acting anticholinesterasic drugs. According to a double-blind placebo-controlled study, drugs with cholinergic action may influence REM sleep [40].

Moreover, a possible effect is related to changes in respiratory parameters in AD patients with obstructive sleep apnea. Indeed, cholinergic activity influences the functional status of the upper airways through central and peripheral mechanisms [41]. This aspect should be considered as a potential confounding factor for the possible action of this type of treatment on the sleep profile. However, it is essential to point out that pharmacological activation of central cholinergic systems may improve abnormalities in sleep patterns [42]. Therefore, it is likely that the negative changes in sleep actigraphic parameters in our AD patients can be considered independent of the drug’s influence. Some of the limitations mentioned above, especially the low sample size, can probably explain some of our results that seem to be at odds with previously reported findings. Our data obtained in MCI did not confirm a reduction in TST [43].

Furthermore, we could not confirm an influence of APOE genotype on sleep patterns [23]. As a further limitation of our study, we did not obtain data on APOE4 genotype in control subjects. On the other hand, our data confirm that actigraphy is a reliable approach to detect circadian rhythm alteration [44, 45].

Actigraphy is not suitable for a detailed description of sleep’s macro and microstructure and defines some particular alterations, including reducing slow sleep waves that characterize neurodegenerative changes. For this reason, a detailed sleep assessment capable of supporting strategic information to define the relationship between sleep and cognitive decline requires polysomnographic evaluation. The selection of patients with sleep disorders, especially in the early phase of cognitive impairment, may be relevant for an early attempt to precisely correct alterations potentially interfering with the evolution of cognitive decline. In this regard, reduced sleep efficiency may be linked to several specific alterations that require a differentiated pharmacological approach. For example, an increase in SL can be corrected by a chronobiotic treatment, while a sleep stabilizer can correct an increase in fragmentation. The possibility of obtaining early objective information with a simple and non-invasive approach may be relevant and expand the potential use of actigraphy to monitor the evolution of sleep disorders with repeated recordings.

The relevance of early assessment and correction of risk factors in the management of cognitive impairment suggests the potential importance of careful assessment of sleep characteristics in patients with cognitive impairment. According to our results, actigraphy should be considered for a simple and non-invasive approach capable of supporting adequate information to guide specific therapeutic approaches and to select patients in whom more specific diagnostic assessments are needed.

AB, amyloid; AD, Alzheimer’s disease; ADL, Activities of Daily Living; CS, control subjects; IADL, Instrumental Activities of Daily Living; MA, mean motor activity; MCI, mild cognitive impairment; MMSE, Mini-Mental Status Examination; MRI, magnetic resonance imaging; NA, number of awakenings; PSQI, Pittsburgh Sleep Quality Index; SE, sleep efficiency; SL, sleep latency; TIB, time in bed; TST, total sleep time; WASO, wakefulness after sleep onset.

LB: drafting/revising the manuscript, study concept or design, analysis and interpretation of data; RC, AP, CR, GV, SB, CF: drafting/revising the manuscript, acquisition of data; LF: drafting/revising the manuscript, analysis and interpretation of data; MS: drafting/revising the manuscript, study concept or design, analysis or interpretation of data, study supervision. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript.

All participants provided written informed consent. The Institutional Review Board of the Department of Experimental and Clinical Medicine, Marche Polytechnic University, approved the study, code 2019/37.

We thank Viviana Totaro for her technical assistance.

This research was funded by the Department of Experimental and Clinical Medicine, Marche Polytechnic University (Research Activity 2019), grant number DMSC/19.

The authors declare no conflict of interest.