Anxious–depressive symptoms and sleep disturbances across the Alzheimer disease spectrum

Abstract

Patients with Alzheimer disease (AD) often experience neuropsychiatric symptoms, particularly anxious–depressive symptoms and sleep disturbances. These symptoms are associated with various factors related to AD, including amyloid-β and tau pathology, neurodegeneration, and cognitive decline, at different stages of the disease. However, it remains unclear whether anxious–depressive symptoms and sleep disturbances are merely symptoms or contribute as risk factors in the development and progression of AD. Consequently, there is a pressing need for a timely and informed discussion regarding these disturbances in AD. Here we discuss the most recent developments in understanding the etiology of anxious–depressive symptoms and sleep disturbances in AD, with a focus on how these symptoms interact with AD biomarkers to influence cognitive decline. Furthermore, we propose models of connections between anxious–depressive symptoms and/or sleep disturbances, AD biomarkers and cognition, aiming to inspire potential treatment plans for addressing these symptoms and exploring their impact on AD pathology and cognitive decline.

Alzheimer disease (AD) is a progressive neurodegenerative disorder and the main cause of dementia, affecting tens of millions of people worldwide. Despite intensive efforts, the cause of AD remains poorly understood, and its early diagnosis and treatment are still challenging. Biologically, AD is defined by the abnormal accumulation of extracellular amyloid-β (Aβ)-containing plaques, intracellular tau-containing neurofibrillary tangles and neurodegeneration^1,2. According to the Aβ/tau/(neurodegeneration) (AT(N)) research framework (Box 1)^1,3, the Aβ and tau biomarkers are used to define AD and related disorders in vivo, and the neurodegeneration biomarkers are used to clinically stage the disease. Clinically, AD is characterized by multifaceted symptoms, including cognitive decline and neuropsychiatric or behavioral manifestations (neuropsychiatric symptoms (NPSs); Box 2), ultimately leading to severe functional disability.

BOX 1. AT(N) framework for AD diagnosis.

The ATN classification system represents a significant advancement in the diagnostic framework for AD, shifting the focus from clinical syndromes to biological constructs. This system categorizes individuals on the basis of biomarker evidence of pathology, which enhances the precision of AD diagnosis and research. This framework evaluates three primary biomarkers: β-amyloid plaques, hyperphosphorylated tau and neurodegeneration. β-amyloid deposition is detected using amyloid PET, CSF Aβ₄₂ or plasma Aβ_42/40 and is categorized as ‘A.’ Hyperphosphorylated tau is measured through tau-PET, CSF p-tau or plasma p-tau and is denoted as ‘T.’ Notably, p-tau₁₈₁ is more sensitive to Aβ pathology, as indicated by amyloid PET, while p-tau₂₁₇ is more closely linked to tau pathology, as measured by tau-PET. Neurodegeneration is assessed via structural MRI for atrophy, FDG PET for glucose metabolism or CSF/plasma t-tau, and is indicated as ‘N’^1,199,200. It is important to note that while CSF/plasma t-tau is often considered a marker of neurodegeneration, it is less specific to AD than other neurodegeneration measures, such as MRI or FDG PET.

Each of the three measures for amyloid and tau-PET, CSF and plasma offers distinct advantages and disadvantages. PET imaging provides a direct in vivo visualization of amyloid and tau pathology, offering high spatial resolution and enabling researchers to observe the regional distribution of these proteins in the brain. However, PET scans are costly, require specialized equipment and involve radiation exposure, making them less accessible for large-scale or repeated use. CSF analysis is a highly sensitive and specific method for detecting amyloid and tau levels, particularly for early detection. However, CSF collection is invasive, requiring lumbar puncture, which can be uncomfortable for patients. Plasma-based biomarkers offer a less-invasive, cost-effective and easily accessible alternative for large-scale screening and monitoring of amyloid and tau. Although plasma measures show promise, they are currently less sensitive and reliable than PET and CSF as they are influenced by peripheral factors and the blood–brain barrier, making them less specific to brain pathology. Each method has trade-offs, with PET and CSF offering high accuracy but being more invasive or costly, while plasma biomarkers hold promise for large-scale use but require further validation.

The 2024 criteria for AD diagnosis³ update the 2018 framework¹ by emphasizing Aβ and tau biomarkers, integrating biological markers with clinical symptoms to better reflect the disease continuum and recent scientific advances. New categories of biomarkers, including plasma biomarkers and later-changing biomarkers, improve diagnostic precision and prognosis. The inclusion of vascular brain injury (V; for example, MRI-measured white-matter hyperintensities volume, covert brain infarcts and free-water fraction) and inflammation (I) biomarkers further captures the multifaceted nature of AD. Research shows a strong link between increased inflammation, ADS and SD in CN elderly individuals. Plasma data revealed that higher baseline levels of inflammatory markers such as IL-6, IL-8 and CRP were associated with increased depressive symptoms over time^193,194, suggesting that inflammation promotes depression. Similarly, poor sleep quality and daytime sleepiness were associated with higher levels of inflammatory markers such as monocyte chemoattractant protein-1 (MCP-1) and chitinase-3-like protein 1 (YKL-40) in CSF^195,196, indicating greater AD pathology. In addition, two studies have highlighted the relationship between prolonged sleep duration and elevated levels of inflammatory markers in CN middle-aged¹⁹⁸ and elderly¹⁹⁷ individuals with a parental history of AD. Reference 197 found that CN elderly individuals nearing or surpassing the age at which their parent developed AD dementia exhibited longer sleep durations, which correlated with higher levels of inflammatory biomarkers in CSF, such as IL-6 and MCP-1. These findings suggest that prolonged sleep in CN older individuals may act as a compensatory mechanism triggered by inflammation as they approach AD onset, potentially helping to mitigate early neurodegenerative changes.

BOX 2. Cognitive deficits and neuropsychiatric symptoms in AD.

Cognitive dysfunction is a hallmark of AD, manifesting through various impairments that progressively worsen over time. One of the most common and early AD signs is memory loss, particularly affecting recent events and new information acquisition. As AD progresses, patients often experience language deficits, including difficulty finding words, reduced vocabulary and challenges in following conversations. Executive control functions, which encompass planning, decision-making and problem-solving abilities, also deteriorate, leading to difficulties in managing daily tasks and maintaining independence. In addition, visuospatial skills decline, making it hard for individuals to navigate familiar environments or recognize faces. These cognitive impairments not only are debilitating for patients but also impose significant emotional and logistical burdens on caregivers^1,201.

NPSs often emerge early in the AD process²⁰², significantly affecting patients’ quality of life and caregiver burden. These symptoms include a wide range of manifestations such as depression, anxiety, apathy, agitation, aggression, psychosis and sleep disturbances. NPSs are prevalent across all stages of AD and often exacerbate as the disease progresses, complicating the management and treatment strategies. The presence of NPSs in AD patients has been linked to faster cognitive decline, increased functional impairment and greater caregiver burden²⁰³. Addressing these symptoms through a combination of pharmacological and non-pharmacological interventions is crucial for improving a patient’s overall quality of life and reducing the strain on caregivers²⁰².

Sleep disturbances and anxious–depressive symptoms are among the most common NPSs to precede cognitive diagnoses of MCI or dementia, significantly impacting the quality of life and the progression of the disease^8,1,6. Patients with AD often experience various sleep disturbances, including short sleep duration, low sleep efficiency and disrupted SWS²⁰⁴. These sleep abnormalities can exacerbate cognitive decline and are linked to the accumulation of β-amyloid plaques, a hallmark of AD. Anxious–depressive symptoms, characterized by sadness, anxiety and a pervasive sense of emptiness, are commonly observed in AD patients. These symptoms not only exacerbate cognitive decline but also complicate the management of the AD, underscoring the importance of addressing these issues in patient care.

Commonly used measures for NPSs and cognitive decline

NPS measures		Cognitive measures
ADS	SD
Neuropsychiatric Inventory Questionnaire (NPIQ)	Pittsburgh Sleep Quality Index (PSQI)	Mini-Mental State Examination (MMSE)
Mild Behavioral Impairment−Checklist (MBI-C)	Actigraphy-derived sleep parameters	PrecLinical Alzheimer Cognitive Composite (PACC)
Beck Anxiety Inventories (BAI)	Electroencephalography/polysomnography−derived sleep waves	Rey Auditory Verbal Learning Test (RAVLT)
Geriatric Anxiety Scale (GAS)	Sleep logs	Clinical Dementia Rating (CDR)
Hospital Anxiety and Depression Scale−Anxiety (HADS-A)		Alzheimer’s Disease Assessment Scale (ADAS)
Beck Depression Inventories–second edition (BDI-II)		Subjective Cognitive Decline Questionnaire (SCD-Q)
Geriatric Depression Scale (GDS)		Cognitive Change Index (CCI)
Hospital Anxiety and Depression Scale−Depression (HADS-D)		Neuropsychological tests
Hamilton Depression Rating Scale (HAM-D)
Center for Epidemiologic Studies Depression Scale (CES-D)
Profile of Mood States−Short Form (POMS-SF)

Late-life NPSs are commonly present in AD and encompass a broad spectrum of conditions, such as anxious–depressive symptoms (ADS), sleep disturbances (SD), elation/euphoria, apathy, agitation, disinhibition, motor disturbance, appetite and eating disturbances, and psychosis⁴. These symptoms not only affect patients’ quality of life but also complicate the clinical management of AD. Addressing these symptoms can improve daily functioning and overall well-being. Of note, NPSs can sometimes be the earliest or most prominent manifestation of AD, even before memory problems become apparent. Moreover, NPSs increased the risk of incident mild cognitive impairment (MCI) to a higher degree than hippocampal atrophy (measured by structural magnetic resonance imaging (MRI)), affirming the clinical relevance of NPSs compared with other well-established predictors of conversion from MCI to dementia⁵. In addition, there is growing evidence that NPSs in AD may be associated with faster cognitive decline and greater functional impairment^6–13. These findings provide evidence that NPSs, when appropriately measured and operationalized, may facilitate early detection of AD and optimized interventions in early stages. However, it is currently uncertain which factors contribute to the emergence of NPSs in AD or whether NPSs contribute to AD development (Box 3).

BOX 3. Stages of AD.

AD typically progresses through several stages, marked by increasing severity of cognitive decline. The initial stage, preclinical AD, is characterized by the presence of pathological changes in the brain, such as the accumulation of amyloid-β plaques and tau tangles, without noticeable symptoms (for example, CN, SCD)^1,205. Following this stage is prodromal AD (MCI due to AD), where individuals begin to exhibit minor but noticeable memory problems and cognitive deficits that do not significantly interfere with daily activities²⁰¹. As the disease progresses, AD dementia manifests in several stages: mild dementia due to AD, where individuals experience memory loss and other cognitive difficulties that begin to impact daily living; moderate dementia due to AD, marked by a more pronounced decline in memory and cognitive functions, significant confusion and a greater need for assistance with daily activities; and severe dementia due to AD, characterized by profound cognitive decline, loss of ability to communicate and complete dependence on caregivers for basic needs^206–208. Understanding these stages is crucial for timely diagnosis and intervention, potentially slowing disease progression and improving quality of life for those affected.

Recent studies have increasingly emphasized the very early stages of AD, particularly the SCD stage. SCD is a self-reported, persistent decline in cognitive abilities, typically involving memory but potentially affecting attention or executive functioning, despite normal performance on cognitive tests and no link to acute events²⁰⁹. According to the National Institute on Aging-Alzheimer’s Association research framework, SCD is classified as Stage 2 of preclinical AD²¹⁰, reflecting its significance as an early indicator of underlying neuropathology and its potential for early intervention. Studies have reported that approximately 14% individuals with SCD develop dementia, and 27% progress to MCI within a decade²⁰⁶. The risk of progression to AD increases when SCD co-occurs with SCD-plus features, such as AD pathologies (Aβ and tau), APOEε4 genotype, memory-focused complaints and mild behavioral impairment^{6,209,211,212}.

SCD is strongly linked to early abnormalities in AD biomarkers, such as Aβ and tau pathologic accumulations, and brain atrophy, underscoring its importance for identifying individuals at higher risk of disease progression. Individuals with SCD who exhibit AD biomarker profiles, such as Aβ pathology combined with tau or neurodegeneration, were at a significantly increased risk of progressing to MCI or dementia compared with biomarker-negative individuals²¹². Furthermore, the combination of Aβ positivity with tau and/or neurodegeneration biomarkers (for example, A+T−N−, A+T+N−, A+T+N+) was associated with steeper cognitive decline and increased risk of dementia²¹¹. Among neurodegeneration markers, hippocampal volume, serum neurofilament light and serum glial fibrillary acidic protein independently predicted clinical progression to MCI or dementia in individuals with SCD, beyond the effects of Aβ and tau²¹³. Another structural MRI study²¹⁴ in SCD populations revealed subtle but significant patterns of brain atrophy, including medial temporal lobe atrophy and hippocampal volume loss, which were similar to those seen in early AD and predictive of conversion to MCI or dementia.

SCD serves as a pivotal early window for detecting AD-related pathology. Integrating multimodal biomarker assessments—encompassing Aβ, tau and neurodegeneration (for example, measured by MRI) measures—improves the identification of high-risk individuals and provides opportunities for timely intervention during this preclinical stage of AD.

In this Perspective, we focus on ADS and SD due to their prevalence and significance during the progression of AD^14–16. ADS in AD patients, such as persistent anxiety, fear and sadness, and loss of interest in activities are well documented^15,17. SD in AD patients manifest as disrupted sleep continuity and architecture and an altered sleep–wake cycle^18,19. We first review the existing empirical data on the dynamic relationships and interactive effects between ADS/SD, AT(N) biomarkers and cognitive performance across different stages of AD (especially preclinical and prodromal AD). We then propose models that provide frameworks for understanding the relationships between ADS, SD, AT(N) biomarkers and cognition. Finally, we discuss potential treatment plans targeting ADS and/or SD and their impact on AT(N) and cognition. In particular, we highlight the advantages and uniqueness of each treatment concerning the modulation of ADS/SD symptoms, AT(N) biomarkers or cognitive performance.

Relationship between AT(N) and ADS

Relationship between Aβ and ADS

Aβ measured using PET imaging.

Multiple studies have found significant associations between Aβ pathology and ADS (Table 1). For example, a study measured cortical Aβ burden in elderly individuals and found that Aβ-positive (Aβ+) individuals with MCI had an increased risk of developing NPSs, particularly ADS and SD, compared with Aβ-negative (Aβ−) individuals with MCI or cognitively normal (CN) individuals²⁰. Another study found that Aβ+ elderly individuals with MCI with depressive symptoms had an elevated Aβ load in frontotemporal and insular cortices compared with those without depressive symptoms²¹. Other studies similarly linked higher Aβ load to ADS in samples with mixed Aβ status. For example, studies involving MCI and CN older adults with mixed Aβ status showed elevated Aβ deposition in cortical and subcortical (the amygdala, striatum and thalamus) regions associating with increased ADS^22,23. In another study with CN adults at risk of AD (that is, with a parental or multiple-sibling family history of sporadic AD and with autosomal dominant AD), cortical Aβ deposition correlated with anxiety, apathy and overall neuropsychiatric score²⁴. These findings indicate that Aβ deposition is intricately linked to the presence and severity of ADS in individuals at various stages of cognitive impairment, underscoring the critical role of Aβ pathology in the neuropsychiatric manifestations of AD.

Table 1 |.

Relationship between ADS, AD pathologies and cognitive decline at different AD stages

Pathology	Study design	AD stage
		Preclinical AD	Prodromal AD	AD dementia
Aβ pathology (A)	Cross-sectional association	PET/CSF: Aβ deposition associates with ADS;22–24,31,35,39 plasma: high41 or low43,44,191 Aβ burden associates with ADS	PET/CSF: Aβ deposition associates with ADS;20–23,31,36 plasma: low Aβ burden associates with ADS191	No data
	Longitudinal association	PET: Aβ deposition predicts ADS increases;25–27 Aβ increases associate with ADS increases;29 CSF/plasma: high37,42 or low38,41 Aβ burden predicts ADS increases	No data	No data
	Association with cognition	PET: Aβ+/ADS+ predict cognitive decline;23,31–33 CSF: Aβ+/ADS+39 or Aβ–/ADS+40 predict cognitive decline; plasma: Aβ–/ADS+ predict cognitive decline44	PET: Aβ+/ADS+ predict cognitive decline;21,23,30 no CSF/plasma data	No data
Tau pathology (T)	Cross-sectional association	PET/CSF/plasma: tau accumulation associates with ADS24,31,47,48,52	PET/CSF: tau accumulation associates with ADS;31,50 no plasma data	PET/plasma: tau accumulation associates with ADS;50,53 no CSF data
	Longitudinal association	No data	No data	No data
	Association with cognition	PET: tau+/depressive-symptoms+ predict memory and executive decline;49 no CSF/plasma data	No data	No data
Glucose metabolism (N)	Cross-sectional association	ADS associate with hypometabolism54–56	Depressive symptoms associate with hypermetabolism21	ADS associate with hypermetabolism58
	Longitudinal association	No data	No data	No data
	Association with cognition	Hypometabolism/ADS+ predict cognitive decline57	No data	No data
Brain atrophy/activity/connectivity (N)	Cross-sectional association	Depressive symptoms associate with lower GMV54,55, reduced white-matter integrity55 and increased anterior DMN connectivity66	Depressive symptoms associate with altered brain activity192 and amygdala connectivity65	Depressive symptoms associate with decreased amygdala–DMN connectivity64
	Longitudinal association	No data	ADS predict greater atrophy in AD-related regions30,59,60	No data
	Association with cognition	No data	No data	Amygdala–DMN connectivity mediates the link between depressive symptoms and cognition64
Vascular brain injury (V)	Cross-sectional association	Positive61,62 or no63 associations between WMH burden and ADS	High WMH burden associates with ADS62	No data
	Longitudinal association	No data	No data	No data
	Association with cognition	WMH burden does not moderate associations between depressive symptoms and cognitive decline63	No data	No data
Inflammation (I)	Cross-sectional association	Plasma: IL-6 associates with depressive symptoms;193 no CSF data	No data	No data
	Longitudinal association	Plasma: IL-6, IL-8 and CRP predict depressive symptoms;193,194 no CSF data	No data	No data
	Association with cognition	No data	No data	No data

IL-6, interleukin-6; IL-8, interleukin-8; CRP, C-reactive protein; A, amyloid-β pathology; T, tau pathology; N, neurodegeneration; V, vascular brain injury; I, inflammation.

Longitudinal evidence suggests that high baseline Aβ deposition is a predictor of ADS progression in CN older adults. For example, one study found that depressive and apathy–anxiety symptoms increased over 6 years in cortical Aβ+ CN older individuals at baseline, compared with Aβ− individuals²⁵. Other studies with samples of mixed Aβ status also reported dynamic associations between baseline cortical Aβ levels and increased depressive symptoms²⁶, ADS²⁷ and incidence of clinical depression²⁸. A recent study further explored the relationships between changes in Aβ, ADS and cognition over time, finding that increasing depressive symptoms over an average of 8.6 years were associated with both increasing Aβ accumulation in frontal and cingulate cortices and declining cognitive function in CN older adults²⁹.

A large body of literature suggests an interaction between ADS and Aβ status in the way they predict cognitive decline. For example, MCI and CN elderly individuals with both Aβ pathology and ADS exhibited faster and more pronounced cognitive decline over time compared with those with either factor alone or neither^{21,23,30–32}. In another study, Aβ+ CN elderly individuals with NPSs had faster global cognitive decline over a median of 6.2 years compared with Aβ+ and Aβ−/NPSs− individuals³³. In the same study, Aβ pathology and multiple NPSs domains (for example, anxiety, depression, irritability, motor behavior, agitation, appetite change, euphoria and irritability) interactively accelerated global and domain-specific cognitive decline (for example, memory, attention, language and visuospatial skills). Moreover, a longitudinal Aβ positron emission tomography (PET) study during AD progression showed that striatal Aβ deposition that followed neocortical Aβ deposition predicted rapid cognitive decline in CN individuals, MCI and AD dementia³⁴. In addition, Aβ levels in subcortical regions (striatum, amygdala and thalamus) were correlated with greater anxiety among CN older adults, particularly in apolipoprotein E4 (APOEε4) carriers³⁵. This study highlighted the role of anxiety as a behavioral indicator in late preclinical AD and suggested that increased anxiety symptoms, coupled with high-risk biological factors such as APOEε4 and advanced Aβ stage, may enhance risk for progression to MCI or AD dementia.

In summary, PET data provide converging evidence that Aβ pathology is associated with ADS at different AD stages.

Aβ measured as cerebrospinal fluid Aβ₄₂.

Some studies have found that lower levels of cerebrospinal fluid (CSF) Aβ₄₂ (indicating higher AD pathology) are associated with more severe ADS, while others have shown the opposite pattern. In CN and MCI older individuals, lower Aβ₄₂ levels were associated with ADS^31,36. Similarly, ref. 37 found that Aβ+ CN elderly developed more anxiety, anger and fatigue over an average of 3 years compared with Aβ− individuals. However, according to ref. 38, a higher Aβ₄₂/Aβ₄₀ ratio in Aβ+ CN elderly was correlated with increased anxiety and apathy over 8 years, a pattern not seen in Aβ− individuals. The discrepancy could be due to several factors such as the different definitions of Aβ positivity, differences in severity of ADS, the inconsistent neuropsychiatric measures across studies or the study duration.

Evidence suggests that CSF Aβ biomarkers can modulate the associations between depressive symptoms, cognitive impairments and AD risk. For example, a study reported that stronger baseline depressive symptoms were linked to higher CSF Aβ burden and poorer global cognition in Aβ+ CN elderly individuals³⁹. Moreover, depressive symptoms and amyloid pathology exacerbated each other over time, leading to an 83% higher risk of developing AD dementia over an average period of 2.9 years. Similarly, another study showed that higher baseline depressive symptoms predicted faster declines in CSF Aβ₄₂, increases in total tau (t-tau)/Aβ₄₂ and progression to MCI over an average of 12.7 years in middle-aged CN adults⁴⁰. However, the association between depressive symptoms and progression to MCI was most evident in those with low levels of AD pathology (high Aβ₄₂ and low phosphorylated tau (p-tau)). These findings indicate that depressive symptoms may not only be a consequence of AD pathology but also play a role in accelerating disease progression through complex interactions with disease-related mechanisms that affect Aβ and tau biomarkers. However, it is important to acknowledge the variability in the relationship between Aβ biomarkers and ADS. Lower CSF Aβ₄₂ levels have not been consistently linked to more severe ADS, highlighting discrepancies across studies. This inconsistency underscores the need for further research to clarify the dynamics between Aβ biomarkers and ADS, particularly across different stages of AD progression.

Aβ in plasma.

Findings on the relationship between plasma Aβ levels and (sub)clinical depression are mixed. In CN older adults, higher plasma Aβ₄₀ levels were associated with more depressive symptoms, especially in those who later developed dementia⁴¹. Conversely, those with lower Aβ₄₀ and Aβ₄₂ levels, who did not develop dementia, had a higher risk of developing depressive symptoms over a mean follow-up of 11 years, suggesting that the longitudinal association between low Aβ levels and depressive symptoms is not explained by dementia. Another study of CN older adults found that higher baseline plasma Aβ₄₂ levels were linked to incident depression and AD after 5 years⁴². However, CN older adults with late-life major depression had a lower plasma Aβ₄₂/Aβ₄₀ ratio compared with those without depression^43,44. More severe clinical depression at baseline was also linked to a lower Aβ₄₂/Aβ₄₀ ratio at 1- and 3-year follow-ups⁴³. In addition, those with ‘amyloid-associated depression,’ characterized by both depression and a low plasma Aβ₄₂/Aβ₄₀ ratio, showed greater deficits in memory, visuospatial ability and executive function compared with those without depression⁴⁴. PET studies supported these findings, showing that baseline Aβ deposition predicts depressive symptoms in older adults (CN or with MCI) with mixed Aβ status over time^25–27. However, older adults (CN or with MCI) diagnosed with late-life major depressive disorder had a reduced Aβ burden^45,46. Overall, these findings suggest that Aβ peptides may reveal important insights about the etiology of depression, and their association with depressive mood may vary depending on the stage of depression. Aβ may expedite the onset and progression of depression but decrease upon clinical depression diagnosis. Longitudinal measurements of Aβ are necessary to understand how depressive symptoms evolve with changes in Aβ levels. In addition, it is crucial to explore the role of tau in this relationship, particularly whether clinical depression correlates with tau accumulation as Aβ decreases.

Relationship between tau and ADS

Tau-PET.

Multiple studies have linked tau pathology to ADS across the AD continuum. Reference 47 found that greater ADS correlated with higher tau measured using PET imaging (tau-PET) signals in the entorhinal cortex and hippocampus in Aβ+ CN older adults. Similarly, refs. 24,48 reported associations between tau deposition in the entorhinal cortex and inferior/lateral temporal lobe with ADS in CN older and at-risk individuals with mixed Aβ status. In addition, ADS combined with prefrontal and hippocampal tau burden were found to be associated with memory and executive decline over 2 years in CN middle-aged to older individuals⁴⁹. The link between greater ADS and increased tau burden in entorhinal cortex and precuneus was also observed in MCI and AD dementia⁵⁰.

Tau measured as CSF p-tau.

In Aβ+ CN older adults, (p-tau) levels were associated with ADS⁴⁷. Similarly, in CN and MCI older adults with mixed Aβ status, higher p-tau/Aβ₄₂ ratios were associated with ADS³¹.

Collectively, findings from PET imaging and CSF indicate that tau pathology is linked to ADS across AD stages.

Relationship between neurodegeneration and ADS

Neurodegeneration measured using CSF and plasma t-tau.

Higher CSF t-tau levels were found to associate with greater ADS. For example, MCI elderly with high CSF t-tau, low CSF Aβ₄₂ or both had more anxiety symptoms than those with low CSF t-tau or high CSF Aβ_42.⁵¹ Higher baseline CSF t-tau/Aβ₄₂ ratios predicted greater increases in ADS over 1 and 3 years in CN older adults with mixed Aβ status²⁶ or with Aβ pathology³⁷.

Similarly, elevated plasma t-tau levels and higher t-tau/Aβ₄₂ ratios were associated with more ADS and SD in CN and MCI older individuals with mixed Aβ status^31,52. A recent study⁵³ further found significantly higher plasma t-tau/Aβ₄₂ ratios in AD participants with anxiety compared with those without, but no such differences in plasma biomarkers were observed in CN or MCI individuals.

Neurodegeneration measured using FDG PET.

Several studies using fluorodeoxyglucose (FDG) PET have demonstrated associations between ADS and glucose hypometabolism in frontolimbic regions vulnerable to AD in CN elderly individuals with mixed Aβ status^54,55 and in Aβ+ MCI elderly²¹, with stronger associations in APOEε4 carriers⁵⁶. CN older individuals with both ADS and glucose hypometabolism showed an increased risk of developing incident MCI over a median of 4.8 years⁵⁷. By contrast, ADS in early-onset AD were associated with glucose hypermetabolism in frontolimbic structures⁵⁸.

Neurodegeneration measured using MRI.

Depressive symptoms were found to link to lower gray-matter volume (GMV) in the hippocampus^54,55 and reduced white-matter integrity in the fornix, posterior cingulum and corpus callosum⁵⁵ in CN older adults with mixed Aβ status. In individuals with MCI, ADS have been predictive of greater cortical atrophy^30,59 and white-matter atrophy⁶⁰ in AD-related regions over a span of 2 years. In addition, positive associations between ADS and white-matter hyperintensities (WMH) have been reported in individuals with MCI, subjective cognitive decline (SCD) and CN elderly^61,62, although a study involving CN elderly⁶³ found no such associations.

Beyond brain morphological abnormalities, emerging research emphasizes the role of functional and structural brain networks in understanding the neurobiological basis of ADS across the AD spectrum. In patients with AD⁶⁴ or MCI⁶⁵, those with depressive symptoms showed decreased functional connectivity of the amygdala with key regions of the default mode network (DMN), including the posterior cingulate cortex, middle frontal gyrus, parahippocampal gyrus and medial prefrontal cortex. Amygdala connectivity with the middle frontal gyrus has also been found to mediate the relationship between depressive symptoms and cognitive function⁶⁴. In CN elderly individuals with major depression, increased functional connectivity in the anterior DMN was associated with higher depression levels, while increased connectivity in the posterior DMN correlated with poorer cognitive function⁶⁶. Moreover, anterior DMN connectivity mediated the influence of Aβ on depression severity, suggesting that the dissociation between anterior and posterior DMN connectivity plays a crucial role in linking Aβ pathology to late-life depression during AD progression. In addition, alterations in brain structural network topology may reflect distinct cognitive dysfunction depending on Aβ accumulation in older adults with MCI and major depression⁶⁷. Specifically, disrupted structural network connectivity in the right middle cingulum was associated with impaired recall and recognition in Aβ+ individuals, whereas altered structural network topology in the frontolimbic circuitry influenced attention and executive function in Aβ− individuals. Together, these findings suggest that disruptions in functional and structural brain network connectivity, particularly within the DMN and frontolimbic circuitry, serve as critical pathways linking Aβ pathology, depressive symptoms and cognitive dysfunction across the AD spectrum.

Relationship between cognitive decline and ADS

Several studies have shown that baseline ADS alone can predict the risk of cognitive impairment, independent of AD biomarkers. For example, a study found that specific sets of NPSs at baseline were linked to faster cognitive decline over an average of 4 years in CN elderly individuals⁶⁸. Another study found that baseline depressive symptoms and AD biomarkers (Aβ measured using PET imaging and CSF Aβ₄₂/Aβ₄₀, t-tau/Aβ₄₂, p-tau/Aβ₄₂) independently predicted cognitive decline over an average of 8 years in CN elderly individuals⁶⁹.

Some studies have also examined the differences in ADS across different stages of AD, such as AD, MCI and SCD. These studies found that individuals with SCD had higher levels of ADS compared with those with MCI or AD^70–72. The overall prevalence of depression was 32% in MCI patients⁷³ and 18.53% in AD patients⁷⁴, suggesting that ADS tend to decrease as AD progresses, potentially due to impairments in memory and executive control^75–77.

In summary, ADS consistently correlate with neurodegeneration biomarkers across different stages of AD. Both ADS and specific AD biomarkers independently predict cognitive decline, and their combination appears to be associated with an accelerated rate of decline. These findings underscore the importance of integrating ADS into the assessment and monitoring of AD progression to improve early detection and treatment strategies.

Associations of late-onset and recurrent depressive symptoms with AD biomarkers and risk of AD

Recurrent depressive episodes, particularly those beginning earlier in life, have been consistently linked to Aβ deposition^78,79 and tau accumulation⁸⁰ in AD-associated regions, such as the hippocampus, medial prefrontal cortex, posterior cingulate cortex, entorhinal cortex and amygdala. The cumulative effects of recurrent depressive episodes can lead to pronounced and long-term changes in brain structure and function^78,81, contributing to accelerated cognitive impairment⁸² and an increased risk of AD^81,83, making recurrent depression a significant risk factor for both AD and cognitive decline.

The risk of AD is approximately doubled in individuals with late-onset depressive symptoms compared with those with recurrent depressive symptoms⁸⁴. However, late-onset depressive symptoms are less strongly correlated with Aβ and tau deposition^45,85,86 and are often considered either a symptom of AD or a response to aging-related cognitive decline⁸⁴. Therefore, late-onset depressive episodes may contribute to AD symptoms through neuropathological mechanisms beyond Aβ and tau accumulation, such as vascular changes^84,87 or brain network dysfunction⁸⁸. Longitudinal studies with repeated measurements are needed to determine whether late-onset depressive symptoms represent a response to the progression of cognitive decline or are driven by Aβ and tau accumulation. Furthermore, future research is encouraged to explore differences between late-onset and recurrent episodes of other neuropsychiatric symptoms, such as apathy and agitation, in relation to AD biomarkers and the risk of AD.

Relationship between AT(N) and SD

Relationship between Aβ and SD

Aβ measured using PET imaging.

The associations between Aβ pathology and SD have been demonstrated by multiples studies (Table 2). For example, cortical Aβ+ CN older adults had worse sleep efficiency, longer sleep latency, more nocturnal awakenings, increased daytime sleepiness⁸⁹ and a lower proportion of non-rapid eye movement (NREM) slow-wave activity (SWA, between 0.6 and 1 Hz)⁹⁰ than Aβ− individuals. In the study by ref. 90, older adults with a high tau burden (tau-positive (T+)) within the medial temporal lobe (MTL) exhibited weaker slow oscillation–sleep spindle (SO–SP) coupling than tau-negative (T−) individuals. Interestingly, the Aβ+ and Aβ− groups did not differ in SO–SP coupling impairment, and the T+ and T− groups did not differ in SWA frequencies. These findings highlighted unique associations between sleep features and specific AD pathology: the disruption of SO–SP coupling was unique to MTL tau, while the relative length of 0.6–1.0 Hz SWA was unique to cortical Aβ. In addition, this study identified a significant correlation between MTL tau and cortical Aβ in the whole sample with mixed Aβ status, supporting the amyloid cascade hypothesis of AD⁹¹. According to this hypothesis, tau accumulation in AD typically accelerates only when cortical amyloid is present. Consequently, the association between tau and SO–SP coupling is likely pertinent to AD, surpassing the changes associated with normal aging.

Table 2 |.

Relationship between SD, AD pathologies and cognitive decline at different AD stages

Pathology	Study design	AD stage
		Preclinical AD	Prodromal AD	AD dementia
Aβ pathology (A)	Cross-sectional association	PET: Aβ deposition associates with SD;89,90,92–97 CSF: SD associate with high103 or low106,107 Aβ burden; plasma: SD associate with high110–112 or low108–110 Aβ burden	CSF/plasma: Aβ deposition associates with SD;103,113,114 no PET data	CSF: Aβ deposition associates with SD;103 no PET/plasma data
	Longitudinal association	PET: SD predict Aβ deposition;90,93,98 CSF: SD predict high102 or low104,105 Aβ burden; no plasma data	No data	No data
	Association with cognition	PET: diminished SWA mediates the influence of Aβ burden on impaired memory;92,101 no CSF/plasma data	No data	No data
Tau pathology (T)	Cross-sectional association	PET/CSF: tau accumulation associates with SD;90,94,97,103,115,116 plasma: high52,112 or low108 tau burden associates with SD	PET/CSF: tau accumulation associates with SD;103,116 no plasma data	PET/CSF: tau accumulation associates with SD;103,116 no plasma data
	Longitudinal association	PET: Declining90 or prolonged116 sleep duration predicts tau accumulation; CSF: tau accumulation predicts poor sleep in Aβ+;195 no plasma data	PET: prolonged sleep duration predicts tau accumulation;116 no CSF/plasma data	PET: prolonged sleep duration predicts tau accumulation;116 no CSF/plasma data
	Association with cognition	PET: tau deposition may mediate the relation between prolonged sleep duration and memory dysfunction;116 no CSF/plasma data	PET: tau deposition may mediate the relation between prolonged sleep duration and memory dysfunction;116 no CSF/plasma data	PET: tau deposition may mediate the relation between prolonged sleep duration and memory dysfunction;116 no CSF/plasma data
Glucose metabolism (N)	Cross-sectional association	SD associate with hypermetabolism123 or hypometabolism95,106	SD associate with hypometabolism95	No data
	Longitudinal association	SD predict hypometabolism123	No data	No data
	Association with cognition	Sleep fragmentation mediates association between fronto-hippocampal hypometabolism and lower executive function95	No data	No data
Brain atrophy/activity/connectivity (N)	Cross-sectional association	SD associate with reduced GMV and cortical thickness95,96,106,124,125,128 and with altered brain activity131 and connectivity129,130	SD associate with altered brain activity131 and connectivity130	SD associate with altered brain activity131 and connectivity130
	Longitudinal association	SD predict GMV decline124	No data	No data
	Association with cognition	No data	Sleep efficiency mediates the link between brain activity and cognitive processing speed131	Sleep efficiency mediates the link between brain activity and cognitive processing speed131
Vascular brain injury (V)	Cross-sectional association	No data	No data	No data
	Longitudinal association	Extensive increases in sleep duration by ≥2hours over time (while still within a total of 7–8h) predict higher WMH volumes and greater FW fraction127	No data	No data
	Association with cognition	No data	No data	No data
Inflammation (I)	Cross-sectional association	CSF: SD associate with MCP-1, MCP-1/Aβ42, YKL-40/Aβ42 and IL-6;196,197 plasma: SD associate with IL-10198	No data	No data
	Longitudinal association	CSF: YKL-40 predicts poor sleep in Aβ+;195 no plasma data	No data	No data
	Association with cognition	No data	No data	No data

FW, free-water; MCP-1, monocyte chemoattractant protein-1; YKL-40, chitinase-3-like protein 1; IL-10, interleukin-10.

In CN middle-aged and older adults with mixed Aβ status, cortical Aβ burden was associated with decreased SWA at 0.6–2.0:Hz^90,92–94, increased sleep fragmentation⁹⁵, longer sleep latency⁹⁶ and worse sleep quality⁹⁷. Longitudinal studies in CN older adults have shown that reduced baseline SWA at 0.6–1.0 Hz⁹³, lower sleep efficiency⁹³ and shorter sleep duration during 40s to 70s⁹⁰ were predictive of increased cortical Aβ deposition later in life. Notably, one night of sleep deprivation led to significant Aβ increases in the right hippocampus and thalamus in CN adults⁹⁸. This study also showed that shorter sleep duration at baseline was associated with higher Aβ burden in subcortical regions. In mice, as in humans, sleep deprivation increased Aβ deposition in a TREM2- (Triggering Receptor Expressed on Myeloid Cells–2) dependent manner⁹⁹ and accelerated Aβ deposition and Aβ plaque-associated tau seeding and spreading in an APOEε4-dependent manner¹⁰⁰.

Some studies suggest that NREM SWA can modulate the influence of Aβ burden on cognitive function in CN elderly individuals. For example, one study found significant associations between reduced SWA, Aβ accumulation in the medial prefrontal cortex and impaired hippocampus-dependent memory consolidation and transformation in a sample with mixed Aβ status⁹². Moreover, the connection between Aβ burden in the medial prefrontal cortex and impaired memory consolidation was mediated by diminished SWA. A subsequent study from the same cohort further demonstrated the role of SWA in moderating the impact of Aβ status on memory function¹⁰¹. This study found that SWA played a crucial role in supporting better memory function, particularly in Aβ+ individuals, who were most in need of cognitive reserve. However, Aβ− individuals, who had less need for cognitive reserve, did not experience the same memory benefits from SWA. Findings from this study suggest that sleep disruption may facilitate the effect of Aβ pathology on cognitive decline in older adults. In addition, SWA might act as a novel cognitive reserve factor, offering protection against memory impairment linked to a significant burden of AD pathology.

In summary, findings from PET studies indicate that increased Aβ burden is consistently associated with disrupted sleep patterns, particularly SWA, and these disruptions may exacerbate cognitive decline through impaired memory consolidation and other mechanisms. This suggests that interventions targeting SWA may hold promise for mitigating the progression and cognitive impacts of AD.

Aβ measured as CSF Aβ₄₂.

Several studies have shown a link between SD and low CSF Aβ₄₂ levels (indicating high AD pathology). For example, Aβ+ CN middle-aged adults exhibited worse sleep efficiency, more nocturnal awakenings and increased daytime sleepiness over 2 weeks compared with Aβ− individuals¹⁰². In another study, reduced CSF Aβ₄₂ levels were linked to worse sleep efficiency, shorter sleep duration, less NREM stage 3 sleep and less rapid eye movement (REM) sleep across various stages of cognitive impairment, including CN, SCD, MCI, mild AD and moderate to severe AD¹⁰³. This study also found that SD occurred earlier than cognitive decline, with both sleep (including NREM and REM sleep) and memory functions worsening as AD progressed.

However, different results have been found in studies involving CN individuals with mixed Aβ status. For example, in middle-aged and elderly individuals, both sleep deprivation¹⁰⁴ and disruption of SWA at 0.5–4.0 Hz¹⁰⁵ increased Aβ₄₂ and Aβ₄₀ levels the following morning, while one night of unrestricted sleep reversed this increase¹⁰⁴. Similarly, other studies found that higher Aβ₄₂ levels were associated with shorter sleep duration¹⁰⁴, worse sleep quality¹⁰⁶ and lower spindle density during NREM stage 2 sleep¹⁰⁷ in middle-aged and elderly individuals. Overall, the relationship between SD and CSF-measured Aβ pathology in AD varies depending on disease stage, Aβ status and specific sleep parameters being measured.

Aβ in plasma.

The relationship between SD and plasma Aβ levels also varies across AD stages and specific sleep parameters. In CN middle-aged adults, sleep deprivation has been shown to decrease Aβ₄₀ and Aβ₄₂ levels^108,109. In CN younger and older adults, slow-wave oscillations¹¹⁰, short sleep duration and low sleep efficiency¹¹¹ were associated with higher Aβ₄₀ and Aβ₄₂ levels in plasma, whereas REM sleep was related to decreased Aβ₄₀ and Aβ₄₂ levels¹¹⁰. In Aβ+ CN older adults, multiple impaired sleep characteristics was associated with higher levels of Aβ₄₂ and Aβ₄₂/Aβ_40.¹¹² In older adults with amnestic MCI, elevated levels of Aβ₄₂ were found to be associated with disrupted slow-wave sleep (SWS)¹¹³ and poor sleep quality¹¹⁴.

Relationship between tau and SD

Tau measured using PET imaging.

Both animal and human studies have demonstrated associations between SD and tau pathology. For example, a mice study found that sleep deprivation increased Aβ plaque-associated tau seeding and spreading in the presence of APOEε4 (ref. 100). In CN older adults with mixed Aβ status, increased tauopathy in the MTL was associated with decreased 1–2 Hz SWA^94,115, longer sleep duration, poorer sleep quality⁹⁷, more NREM stage 1 sleep¹¹⁵ and increased daytime sleepiness⁹⁴. Another study found that T+ CN elderly individuals had weaker SO–SP coupling than T− individuals⁹⁰. In addition, declining sleep duration in one’s 60s predicted a higher tau burden later in life, regardless of Aβ status. However, prolonged sleep duration in older adults with preclinical AD, prodromal AD or AD dementia was associated with increased tau accumulation in cortical and subcortical regions (for example, insula, anterior cingulate cortex, thalamus and amygdala), and faster disease progression¹¹⁶ seems to contradict earlier observations linking shorter sleep duration and poorer sleep quality to tau pathology. This discrepancy may reflect different stages of disease progression, where shorter sleep is associated with early tau accumulation, while longer sleep duration in later stages could signify compensatory mechanisms or underlying neurodegeneration, further contributing to disease advancement. It underscores the complexity of the relationship between sleep patterns and tau pathology across the AD continuum.

Tau measured as CSF p-tau.

SD have been found to be associated with increased CSF p-tau levels. For example, a study involving individuals with CN, SCI, MCI, mild AD or moderate to severe AD reported that higher levels of p-tau and t-tau were correlated with worse sleep efficiency, shorter sleep duration, less NREM stage 3 sleep and less REM sleep¹⁰³. Another study of CN elderly with mixed Aβ status found that an increased p-tau/Aβ₄₂ ratio was associated with decreased 1–2 Hz SWA⁹⁴. Notably, a study randomized eight Aβ− CN middle-aged adults to either sleep deprivation for 36 hours, increased sleep (especially SWS) with sodium oxybate or normal sleep conditions and found that sleep loss affected p-tau differently depending on the modified site¹¹⁷. Specifically, sleep deprivation increased phosphorylated tau threonine 181 (pT181) and 217 (pT217), unphosphorylated tau threonine 181 (T181), threonine 217 (T217) and serine 202 (S202), as well as the ratio of pT217/T217, while decreasing the ratio of phosphorylated tau serine 202 (pS202)/unphosphorylated tau serine 202 (S202).

Tau in plasma.

Sleep deprivation decreased plasma pT181, T181 and T217 concentrations in CN middle-aged adults¹⁰⁸. Higher p-tau and t-tau levels were associated with multiple impaired sleep characteristics in CN older adults^52,112.

Overall, evidence from PET, CSF and plasma studies suggests that SD may represent an early sign of tau-related dysfunction, possibly reflecting axonal damage¹¹⁸ or altered neuronal tau secretion¹¹⁹, rendering it a potentially novel marker for early neuronal dysfunction. Given sleep’s critical role in memory consolidation and neuroplasticity, disruptions in key aspects of sleep—such as SWA, sleep spindles, REM sleep and overall sleep efficiency—may provide a pathway through which tau pathology interferes with the brain’s ability to consolidate memories. Specifically, these sleep disturbances could facilitate or exacerbate tau’s detrimental effects on the neural circuits responsible for memory, leading to impaired cognitive function over time.

Relationship between neurodegeneration and SD

Neurodegeneration measured using CSF t-tau.

Numerous studies have linked SD to increased CSF t-tau levels in CN middle-aged and older individuals with mixed Aβ status, although one study involving SCD individuals found no such association¹²⁰. For example, higher t-tau in CSF have been associated with worse sleep efficiency over the preceding six nights¹⁰⁵, lower spindle density during NREM stage 2 sleep, fewer NREM stage 2 spindles and shorter spindle duration¹⁰⁷. Similarly, a higher t-tau/Aβ₄₂ ratio was associated with reduced SWA at 1–2 Hz⁹⁴. The higher t-tau/Aβ₄₂ ratio and lower spindle density were also correlated with poorer cognitive performance^107,121. Notably, a study involving both mice and CN middle-aged humans found that chronic sleep deprivation led to an acute increase in CSF tau levels within hours and promoted tau spreading in the brains of both species¹²².

Neurodegeneration measured using FDG PET.

SD have shown both cross-sectional and longitudinal associations with metabolic dysfunction. In SCD or MCI elderly with mixed Aβ status, increased sleep fragmentation was associated with hypometabolism in the insula⁹⁵. Similarly in CN elderly with mixed Aβ status, poor sleep quality was associated with hypometabolism in the orbitofrontal–limbic–temporal regions¹⁰⁶, whereas sleep fragmentation was associated with hypometabolism in the fronto-hippocampal regions and executive dysfunction, respectively⁹⁵. In ref. 95, sleep fragmentation was further found to mediate the association between fronto-hippocampal hypometabolism and lower executive functioning in CN older adults. In addition, a longitudinal study classified CN elderly into three groups on the basis of the AT(N) framework: preclinical AD (both Aβ and tau present), asymptomatic at risk for AD (either Aβ or tau present) and healthy controls (no Aβ or tau present)¹²³. Findings showed that only in preclinical AD, higher Neuropsychiatric Inventory Questionnaire scores—driven by SD and irritability/lability subdomains—were associated with hypermetabolism in the limbic regions at baseline. These high scores further predicted subsequent hypometabolism in the posterior cingulate cortex over 2 years in preclinical AD.

Neurodegeneration measured using MRI, electroencephalography or MEG.

SD and brain atrophy are closely connected in CN elderly individuals, with different sleep issues affecting specific brain areas. Poor sleep quality, for example, has been linked to reduced GMV in fronto–parieto–limbic regions^106,124,125 and decreased cortical thickness in the mesial frontal cortex¹²⁵. Sleep fragmentation, however, was associated with lower GMV in inferior–frontal–subcortical regions^95,96,125. Increased daytime sleepiness, as well as sleep durations either shorter or longer than 6.5 hours, correlated with reduced GMV in orbitofrontal–parieto–limbic regions^125–127. Moreover, extensive changes in sleep duration—whether increased or decreased—over an average of 13 years had adverse effects on WMH volumes and free-water fraction¹²⁷. In CN late-middle-aged individuals, the proportion of SWS and NREM stage 1 sleep was related to cortical thickness in AD-signature regions¹²⁸. In addition, insomnia symptoms were linked to disrupted functional organization of critical brain networks, such as the DMN, central executive network and salience network (SAN), with distinct connectivity patterns across CN, MCI and AD elderly groups¹²⁹. These findings suggest that distinct sleep disturbances target different brain networks, impacting overall brain health in the aging population.

Interestingly, Aβ status appears to influence how SD affect brain connectivity in older adults. In Aβ+ individuals across the CN, MCI and AD spectrum, SD were linked to hyperconnectivity within the SAN, a pattern not seen in Aβ− individuals¹³⁰. In addition, sleep efficiency had differential effects on brain activity in the delta and alpha frequency bands, as measured by magnetoencephalography (MEG), depending on disease status (that is, AD spectrum versus CN)¹³¹. In individuals along the AD spectrum (that is, amnestic MCI and mild AD), poor sleep efficiency increased delta power and decreased alpha power in the frontotemporal cortices, while the opposite was true for CN individuals. The study also found that sleep efficiency moderated the relationship between cortical Aβ accumulation and delta power deviations, with higher Aβ levels linked to reduced delta power only in individuals with better sleep efficiency. This modulation extended to cognitive performance as sleep efficiency mediated the link between delta activity and cognitive processing speed in AD.

In summary, SD and neurodegeneration, as measured through FDG PET, MRI and MEG, are closely linked across the AD spectrum, with the effects varying on the basis of Aβ status, sleep characteristics and affected brain regions. A key factor in assessing the role of sleep in memory consolidation and neurodegeneration is the frequency range of SWA, typically defined as 0.5–4.0 Hz. However, studies often use slightly different subranges within this spectrum, which can lead to discrepancies in findings and complicates direct comparisons. For example, some studies focus on the 0.6–1.0 Hz range of SWA, linking it to increased Aβ burden, while others examine the 1–2 Hz range to explore its correlation with tau pathology and neurodegeneration. Given that different SWA frequency bands may reflect distinct aspects of sleep physiology, it is critical to standardize the frequency ranges used across studies or to account for these differences in study comparisons. Furthermore, the potential role of SWA in mitigating Aβ and tau pathology may depend on specific sub-bands, underscoring the need for more targeted research to determine which SWA frequency ranges offer the greatest cognitive reserve, particularly in CN individuals or those at risk for AD. Addressing these inconsistencies will improve the generalization of findings and contribute to more effective sleep-based interventions for reducing AD risk.

Relationship between cognitive decline and SD

Several studies have shown a link between SD and cognitive decline, independent of AD biomarkers. For CN older adults, increased night-time wakefulness and decreased sleep efficiency were associated with higher SCD¹³². Another study found that CN adults who consistently slept less at ages 50, 60 and 70 had a 30% higher risk of dementia over an average of 24.6, 14.8 and 7.5 years, respectively¹³³. In a sample with CN or impaired older adults, maintaining a moderate sleep duration (4.5–6.5 hours), time in NREM and REM sleep and <1 Hz SWA, were important for maintaining cognitive function over time¹²¹. In addition, the impact of sleep on cognition varied by AD stage: sleep fragmentation lowered cognitive performance in CN elderly but did not contribute to cognitive deficits in those with SCD or MCI⁹⁵.

Emerging evidence suggests that SD increase progressively from aging to AD. Specifically, AD patients exhibited significantly reduced REM sleep and increased light sleep compared with CN individuals, while MCI patients showed mild reductions in REM and increases in light sleep¹³⁴. Moreover, AD patients reported more SD than both MCI and CN elderly individuals, with those experiencing SD showing greater WMH burden¹³⁵, suggesting that SD may decrease cognition by increasing WMH burden.

Regional effects in the relationship between ADS/SD and AD biomarkers

Neuroimaging studies suggest potential spatial overlaps between the effects of Aβ, tau and ADS/SD on various brain locations. Tau initially accumulates in brain regions within the medial temporal lobe, such as the entorhinal cortex, hippocampus and amygdala, during healthy aging and early AD stages^136,137. As discussed in the section ‘Relationship between tau and ADS‘, tau accumulation in these regions correlates with the severity of ADS in both Aβ+ and Aβ− CN adults^24,47–49. These regions are also known to show abnormal connectivity and/or activity associated with ADS^138–140. Similarly, SD have been linked to tau pathology in medial temporal lobe structures and regions within the SAN, including the insula, anterior cingulate cortex, thalamus and amygdala^{90,94,97,115,116}. The SAN is believed to be involved in detecting salient stimuli and regulating attention. SD are also associated with neurodegeneration biomarkers in the SAN, manifesting as brain atrophy^95,125 and increased functional connectivity^129,130,141.

As described in the sections ‘Relationship between Aβ and ADS‘ and ‘Relationship between Aβ and SD’, current research focused primarily on associations between global cortical Aβ levels and ADS/SD. Among these cortical regions, the DMN appears to play a critical role in studying region-specific associations between Aβ and ADS/SD. The DMN, which includes the medial prefrontal cortex, posterior cingulate cortex, precuneus and inferior parietal lobule^20–29, is involved primarily in self-referential processes^142–144. Evidence shows that Aβ deposition and decreased functional connectivity initially manifest in DMN regions during the progression of AD^145–148. ADS, particularly depressive symptoms, have been linked to reduced gray-matter volume in DMN regions (for example, the hippocampus)^54,55 and increased functional connectivity within the DMN^66,149. Moreover, altered DMN connectivity has been shown to mediate the impact of Aβ burden on depressive symptoms⁶⁶. For SD, studies have consistently reported brain atrophy^106,128 and reduced functional connectivity^150–152 within the DMN. Future studies should investigate region-specific associations between Aβ and ADS/SD, as well as other neuropsychiatric symptom domains.

These findings indicate a potential shared pathway linking ADS/SD, Aβ pathology and neurodegeneration within the DMN while highlighting the medial temporal lobe as a key circuit connecting ADS/SD, tau pathology and neurodegeneration. Importantly, although both Aβ and tau pathologies contribute to ADS/SD, their effects are spatially distinct at initial AD phases: Aβ pathology initially affects regions within the DMN, such as the medial prefrontal cortex and posterior cingulate cortex, whereas early tau pathology is localized predominantly to the medial temporal lobe, including the entorhinal cortex, hippocampus and amygdala. This spatial dissociation underscores the unique contributions of Aβ and tau to distinct neural circuits and their associated neuropsychiatric and cognitive symptoms. The involvement of ADS/SD in both Aβ and tau pathologies across different brain regions suggests that ADS/SD serve as a critical link, bridging the spatially distinct patterns of Aβ and tau accumulation in early AD and mediating their interaction during disease progression.

Hypothetical models

In the preceding sections, we have reviewed evidence demonstrating that both SD and ADS are intricately linked to AD biomarkers in CN individuals and those with MCI. SD and ADS are emerging as early clinical manifestations of AD as they have been found to predict faster cognitive decline. Disruptions in NREM SWA and alterations in sleep architecture likely serve as early indicators of Aβ and tau accumulation, as well as neurodegeneration, and collectively predict cognitive decline. Moreover, Aβ burden has been found to predict the onset of ADS, and together they act synergistically to accelerate cognitive deterioration. ADS are also associated with tau accumulation and neurodegeneration, both of which further exacerbate cognitive impairment. However, the precise temporal relationships between these biomarkers, NPSs and their combined effects on AD progression remain unclear. This area of research is still in its infancy, and several key questions remain unanswered. How do SD and ADS interact with Aβ and tau biomarkers to influence cognitive decline at different stages of AD? Can early interventions targeting SD or ADS alter AD biomarkers and slow disease progression? What is the optimal timing for such interventions to have the greatest impact on delaying cognitive decline and mitigating neurodegeneration? To address these questions, we propose a set of hypothetical models offering theoretical frameworks to study the causal relationships between AT(N) biomarkers, ADS, SD and cognitive decline.

ADS-facilitated AD progression model.

The ADS-facilitated AD progression model (Fig. 1) illustrates interactions between AT(N) biomarkers, ADS and cognitive decline. In this model, ADS occur slightly later than Aβ deposition during late preclinical AD³⁵. Tau accumulation and neurodegeneration emerge after Aβ buildup and correlate with ADS, although the temporal relationships among tau, neurodegeneration and ADS remain unclear. In addition, Aβ interacts with ADS to predict cognitive decline. Future studies could explore the dynamic relationships between Aβ, tau, neurodegeneration and ADS, as well as their effects on cognition. For example, a longitudinal cohort study could be conducted where participants are stratified on the basis of Aβ and tau status (for example, Aβ+/T+, Aβ+/T−, Aβ−/T−). The progression of ADS and neurodegeneration could be tracked over time using neuropsychological assessments and neuroimaging (for example, FDG PET, MRI). This would allow researchers to observe how ADS, Aβ, tau and neurodegeneration interact to predict cognitive decline. Moreover, future studies could group participants into Aβ+ and Aβ− categories and test whether the two groups differ in ADS. If no group differences are found, it would suggest that ADS likely appear earlier than Aβ. Of note, it is important to examine the impact of ADS on the onset and progression of tau pathology, in the presence of Aβ. If treating ADS in Aβ+ individuals leads to a decrease in tau levels, then such treatments (for example, medication, neuromodulation and psychotherapy) in this population might help reduce the progression to symptomatic AD.

Fig. 1 |. ADS-facilitated AD progression model.

This hypothetical model illustrates interactions between AT(N) biomarkers, ADS and cognitive decline. (1) Aβ (for example, potentially in the DMN regions such as the medial prefrontal and posterior cingulate cortices, precuneus and inferior parietal lobule) emerges earlier than ADS during the late preclinical stage of AD and predicts ADS later in life. (2) Tau (for example, in the medial temporal lobe, including the entorhinal cortex and hippocampus) and neurodegeneration (for example, in the DMN regions) emerge later than Aβ and correlate with ADS, although the temporal relationships between tau, neurodegeneration and ADS remain unclear. (3) Aβ interacts with ADS to predict cognitive decline. Specifically, Aβ moderates the influence of ADS on cognition (orange line), while ADS moderate the influence of Aβ on cognition (blue line). (4) Aβ moderates the association between ADS and tau burden.

SD-facilitated AD progression model.

Similarly, there is an urgent need to identify the causal relationship between SD and the onset of AD as this could significantly impact the timing and strategy of therapeutic interventions. Although current evidence supports that the onset of sleep symptoms occurs almost simultaneously with Aβ deposition during preclinical AD, some studies showed that SD starting from mid-life could predict greater late-life cortical Aβ and tau burden^90,93 and higher risk of AD¹³³. The SD-facilitated AD progression model (Fig. 2) illustrates the interactions among AT(N) biomarkers, SD and cognitive decline. In this model, SD emerge early in mid-life, during the early preclinical stage of AD, and predict later-life Aβ accumulation, tau pathology and neurodegeneration, although the temporal order of these biomarkers remains unclear. In addition, SD interact with Aβ and neurodegeneration biomarkers to predict cognitive decline. More longitudinal studies starting from mid-life are needed to clarify the temporal relationships between SD and Aβ accumulation. However, given the inherent challenges of longitudinal studies—such as time, cost, participant dropout and confounding factors—laboratory-based manipulation of sleep may offer a more practical and controlled approach for uncovering causal relationships. One example is the use of sleep deprivation experiments, which have demonstrated their effects on Aβ deposition^98,104,105. Such applied behavior analysis experimental designs offer experimentally tractable opportunities to investigate the effects of both sleep deprivation and subsequent recovery sleep on neuronal and behavioral changes. Evidence has shown that a single night of sleep deprivation can alleviate depressive symptoms in approximately 45% of depressed patients¹⁴¹, while it has a detrimental effect on mood in CN healthy individuals^98,141. However, these mood effects are transient as a single night of recovery sleep can reverse them¹⁴¹. To date, no studies have examined the long-term impact of sleep deprivation on Aβ and tau levels or the effects of recovery sleep on these biomarkers. If sleep deprivation exerts a sustained influence on Aβ and/or tau, it may play a more significant role in the progression of AD than is currently recognized.

Fig. 2 |. SD-facilitated AD progression model.

This hypothetical model illustrates interactions between AT(N) biomarkers, SD and cognitive decline. (1) SD emerge early in mid-life, during the early preclinical stage of AD, and predict later-life Aβ accumulation (for example, potentially in the DMN regions such as the medial prefrontal and posterior cingulate cortices, precuneus and inferior parietal lobule), tau pathology (for example, in the medial temporal lobe regions such as the entorhinal cortex, hippocampus, amygdala and parahippocampus, as well as in brain regions within the SAN, including the insula, anterior cingulate cortex, thalamus and amygdala) and neurodegeneration (for example, in the DMN and SAN regions). The temporal order of these biomarkers remains unclear. (2) SD interact with Aβ and neurodegeneration to predict cognitive decline. Specifically, SD moderate the influence of Aβ burden on cognition (blue line) and the influence of neurodegeneration on cognition (dotted blue line). (3) SD moderate the association between Aβ and neurodegeneration (green line). (4) Aβ moderates the association between SD and tau burden (dotted orange line) as well as between SD and neurodegeneration (orange line).

Temporal progression model.

Data collected over decades could help elucidate the temporal relationships between ADS, SD, AT(N) biomarkers, cognitive decline and AD dementia. Existing research has shown that ADS mediated the relationship between SD and SCD^153,154, while SD have been found to mediate the relationship between ADS and cognitive function in older adults with MCI¹⁵⁵. Figure 3 presents a hypothetical model that outlines one plausible framework for these temporal relationships. We hypothesize that SD begin to manifest in one’s 40s, preceding the appearance of Aβ and serving as an early clinical indicator of AD pathophysiology. ADS typically emerge around one’s 60s and are considered a prodromal feature of AD¹⁷. SD, ADS and Aβ share bidirectional relationships, forming a feedback loop that interactively promotes tau accumulation and neurodegeneration. These pathological changes collectively impair cognitive function, ultimately leading to AD dementia. Although Aβ deposition is positioned early in this model due to our focus on patients with established AD pathology^1,3, ADS, tau and neurodegeneration may occur earlier than Aβ in some cases, acting as independent risk factors or early indicators of cognitive decline. The relationships depicted in this model are inherently complex, involving overlapping and parallel processes rather than a strictly linear sequence. Moreover, the progression of AD is driven by a complex interplay of genetic, neuropathological and environmental factors (possible risk factors are listed in Fig. 3), which can influence both the temporal and causal relationships of these processes. This highlights the dynamic nature of AD progression and the need to view the proposed model as a tentative hypothesis rather than a definitive causal pathway. Future studies are necessary to test these potential models, clarify the temporal and causal relationships among these factors and explore the efficacy of interventions targeting SD and ADS at different stages of AD progression. Such research could assess the ability of these interventions to delay cognitive decline or modulate AD biomarkers. Comparisons between individuals receiving these interventions in mid-life versus later stages of the disease could offer crucial insights into the optimal timing for therapeutic strategies.

Fig. 3 |. Temporal progression model.

This hypothetical model illustrates one plausible framework for the temporal relationships between AT(N) biomarkers, SD, ADS, cognitive decline and AD dementia. (1) SD typically emerge in one’s 40s, earlier than Aβ accumulation, while ADS tend to arise in one’s 60s. SD, ADS and Aβ have bidirectional relationships with each other, resulting in a feedback loop that interactively promotes tau accumulation and neurodegeneration, which in turn affects cognitive function, ultimately leading to AD dementia. While Aβ is positioned in the early stage of this model based on its role as a key initiating event in AD pathology^1,3, SD and ADS may independently influence or interact with tau pathology and neurodegeneration through mechanisms beyond Aβ. The temporal order of these events may vary across individuals, with some processes occurring concurrently or in feedback loops rather than in a strict sequential order. For example, SD and ADS can exacerbate Aβ and tau accumulation while simultaneously being driven by neurodegeneration. In addition, the progression of AD is influenced by a combination of genetic, neuropathological and environmental factors, which can modify both the temporal and causal relationships among these processes. These complexities highlight the need to view the model as a tentative hypothesis rather than a definitive causal pathway. (2) The model suggests potential treatment plans targeting SD and ADS during the preclinical stage of AD to delay disease progression.

Treatments of SD and ADS in AD and future directions

Given the strong association between SD, ADS, AD pathologies and cognitive decline, treating SD and/or ADS at earlier stages may offer a valuable opportunity to delay or alter AD pathogenesis. Evidence suggests that effective non-pharmacological and pharmacological treatments for SD and/or ADS (Fig. 3) could enhance cognitive performance and slow the progression of AD, potentially by modulating AD pathologies. Overall, non-pharmacological interventions, such as psychotherapy and environmental adjustments, should be prioritized, followed by medications with the fewest side effects and for the shortest duration possible¹⁵⁶.

Pharmacological treatments.

Donepezil, an acetylcholinesterase inhibitor commonly used to treat memory impairment in AD, has shown potential in managing sleep disorders, such as obstructive sleep apnea, in patients with or without AD. In addition, donepezil has been associated with improvements in REM sleep¹⁵⁷, cognitive performance¹⁵⁷ and ADS^158,159 in patients with moderate to severe AD. Another medication, melatonin, is used to enhance sleep onset and duration, but its use in patients with dementia has been linked to adverse mood effects. Combining melatonin with bright light therapy (BLT) may help avoid these side effects¹⁶⁰ as studies have demonstrated improved sleep–wake cycles, cognitive function and emotional well-being in patients with MCI^161,162. Suvorexant, the only FDA-approved drug for treating insomnia in individuals with mild to moderate AD, is an orexin receptor antagonist. It has been shown to lower Aβ and tau levels in CN middle-aged adults, suggesting its potential to slow AD progression by regulating the sleep–wake cycle¹⁶³.

For ADS, although there are no FDA-approved medications specifically targeting ADS in AD¹⁶⁴, research indicates that a higher Aβ burden predicts poor antidepressant response in elderly patients with major depressive disorder, suggesting that AD pathology may contribute to treatment resistance¹⁶⁵. Current pharmacotherapies for ADS in AD typically target neurotransmitters such as serotonin and acetylcholine. Selective serotonin reuptake inhibitors have shown efficacy in alleviating ADS in AD patients^166,167, improving psychomotor speed and delayed recall in depressed individuals¹⁶⁸ and reducing Aβ production in both mice and CN adults^169,170. Selective serotonin reuptake inhibitors have also been linked to delayed AD progression in MCI patients with a history of depression when used for extended periods¹⁷¹. Cholinesterase inhibitors, used primarily to improve cognitive function, have also demonstrated efficacy in reducing depressive symptoms in mild to moderate AD¹⁷².

Non-pharmacological treatments.

For SD, BLT has proven effective in improving sleep in AD patients¹⁷³. BLT regulates circadian rhythms and consolidates sleep–wake patterns, leading to improved night-time sleep, reduced nocturnal awakenings and enhanced daytime wakefulness¹⁷⁴. Furthermore, studies suggest that BLT may improve cognitive performance¹⁷⁵. Continuous positive airway pressure therapy for sleep apnea has been shown to improve cognition and delay the onset of dementia, particularly in individuals with sleep apnea, regardless of whether cognitive impairment is present¹⁷⁶. Continuous positive airway pressure thus offers a promising therapeutic approach for mitigating cognitive deficits associated with sleep-related AD pathology.

Psychotherapies, such as cognitive behavioral therapy and reminiscence therapy, have been shown to alleviate ADS^177,178 and enhance cognitive functions^179–181 in AD patients. Other interventions for reducing ADS include sleep restriction¹⁴¹ and transcranial magnetic stimulation (TMS)¹⁸². Acute sleep deprivation has been found to rapidly improve depressive symptoms in patients with major depression¹⁴¹, yet it increases Aβ burden in brain regions vulnerable to early AD (for example, hippocampus and thalamus) in healthy CN individuals⁹⁸. Moreover, sleep deprivation has been shown to worsen mood in healthy CN individuals^98,141, thereby increasing the risk of developing clinical depression¹⁸³. TMS, a non-invasive neuromodulation technique, has been explored as a potential treatment for ADS without inducing Aβ accumulation¹⁸⁴. TMS may improve executive function and attention in depressed patients, although its cognitive effects in AD patients have been inconsistent¹⁸⁵. One promising target for TMS is the DMN, whose connectivity decreases with normal aging, MCI and AD^186,187 and reduces following sleep deprivation¹⁸⁸. By contrast, depression has been associated with increased DMN connectivity^149,189. Future TMS studies could test whether modulating DMN connectivity^182,190 may offer a novel therapeutic approach for AD.

Future directions.

As research continues to evolve, it is essential to determine which phenotypes of SD and ADS require targeted treatment at various stages of AD. Emerging evidence suggests that distinct phenotypes of these symptoms correspond to specific AD pathologies. For example, a reduced proportion of SWA in the 0.6–1.0 Hz range (relative to the broader 0.6–4.0 Hz range) has been uniquely associated with cortical Aβ deposition, while disruptions in SO–SP coupling are linked specifically to tau pathology⁹⁰. Moreover, tau burden has been associated with poorer sleep quality as measured by actigraphy, whereas both Aβ and tau have been linked to worse self-reported sleep quality⁹⁷. Furthermore, anxiety symptoms have been shown to correlate with Aβ deposition²³, while depressive symptoms are more closely associated with tau accumulation⁴⁸. Although the research on phenotype–pathology relationships remains limited, identifying these specific phenotypes of SD and ADS in relation to distinct AD biomarkers is crucial for advancing precision medicine.

Future studies should also explore the differential interactions between ADS, SD and AD-related biomarkers between patients with AD and high-risk individuals who have not yet developed the disease. These studies could help clarify whether ADS and SD in high-risk individuals serve as early indicators of cognitive decline, offering potential biomarkers for early diagnosis and prevention. Moreover, there is an urgent need for longitudinal studies that track high-risk individuals over time to better understand how ADS and SD interact with other AD risk factors, such as lifestyle and family history, in the early stages of disease development. Such research would significantly enhance our understanding of the pathogenesis of AD.

In conclusion, SD may serve as early indicators of AD-related physiological changes, while ADS may represent a later prodrome of AD. Addressing SD and ADS could play a pivotal role in slowing AD progression. The integration of non-pharmacological, pharmacological and neuromodulation therapies, alongside a better understanding of symptom–pathology relationships, will likely enhance therapeutic outcomes. Future studies should prioritize refining diagnostic criteria for neuropsychiatric symptoms in AD, paving the way for more targeted and effective interventions for these prevalent and impactful symptoms. Ultimately, we hope that continued research will facilitate the identification of AD subtypes, address disease heterogeneity and promote the development of more targeted therapeutic strategies tailored specifically to AD-related SD and ADS for modulating disease progression.