Abstract
Introduction
The interplay between glymphatic function and sleep quality is crucial for brain health and cognitive longevity in late adulthood. Beyond chronological age, whether brain age has specific effects on the associations between glymphatic function, sleep quality, and cognition are understudied in cognitive unimpaired adults.
Methods
Structural and diffusion magnetic resonance imaging (MRI) data from the Cambridge Centre for Ageing and Neuroscience (Cam-CAN) project (N = 582, age range 18–87 years) were used to calculate brain age metrics and the diffusion tensor imaging analysis along the perivascular space (DTI-ALPS) index. Brain age metrics comprised estimated brain age and the brain predicted age difference (brain-PAD). Subjective sleep quality was assessed using the Pittsburgh Sleep Quality Index (PSQI). Cognitive assessments included accuracy, reaction time, intraindividual variability of reaction time, and fluid intelligence.
Results
Leftward asymmetry of the DTI-ALPS index was consistently observed across brain age-specific groups. The brain-PAD score was significantly correlated with a lower left DTI-ALPS index. Adults with a positive brain-PAD score exhibited a robust correlation between DTI-ALPS indices and sleep quality features, whereas those with a negative brain-PAD score showed a reliable correlation between DTI-ALPS indices and cognition. Mediation analyses further revealed that the relationship between left DTI-ALPS index and sleep efficiency was mediated by brain age.
Conclusion
This study provides the first demonstration that lateral differences in the DTI-ALPS index vary according to brain ageing statuses. The two distinct profiles of the sleep–glymphatic function–cognition connections observed in relation to brain-PAD scores suggest that a preserved brain age may serve as a protective factor against age-related decline in glymphatic function. These findings may underscore the translational potential of brain age models as both clinical biomarkers and modifiable targets for interventions aimed at promoting healthy longevity and brain resilience.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1007/s40120-026-00999-9.
Keywords: Brain age, DTI-ALPS, Glymphatic function, Sleep quality, Sleep efficiency, Cognition
Key Summary Points
| Why carry out this study? |
| The glymphatic system plays a vital role in maintaining brain health and holds the key to developing mechanism-based personalized gerotherapeutics. The Diffusion Tensor Imaging Along the Perivascular Space (DTI-ALPS) index, as an indirect measure of glymphatic function, declines with chronological age. However, whether brain age differentially affects the associations between DTI-ALPS, sleep quality, and cognition remains unclear. |
| This study aimed to identify brain age-specific patterns of bilateral DTI-ALPS indices and to investigate the relationships among DTI-ALPS indices, sleep quality features, and cognition in a well-characterized cohort of cognitively unimpaired adults. |
| What was learned from the study? |
| This study provides the first evidence that the distinct interplays among sleep quality features, glymphatic function, and cognitive functions across different brain ageing statuses suggest that preserved brain age may serve as a protective factor against age-related decline in glymphatic function. |
| These findings may deepen the understanding of sleep–glymphatic function–cognition connections, shedding new light on the clinical significance of brain age model and its potential as a gerotherapeutic target for enhancing brain resilience and promoting healthy longevity. |
Introduction
The glymphatic system plays a vital role in maintaining brain health by serving as the brain’s primary waste clearance pathway, facilitating the efficient removal of metabolic byproducts, soluble proteins (e.g., amyloid-beta), and other neurotoxic waste through the exchange between cerebrospinal fluid (CSF) and interstitial fluid (ISF) [1]. This process occurs predominantly during sleep and is fundamental for preserving neuronal integrity, preventing protein aggregation, and supporting cognitive longevity [2, 3]. However, glymphatic function undergoes progressive decline with advancing age. Impairments often become noticeable around midlife (between ages of 50 and 55) and accelerating more markedly after age 65 as a result of factors such as depolarization and reduced the polarization of aquaporin-4 (AQP4) channels on astrocytic end feet [4], diminished arterial pulsatility [5], reduced slow-wave sleep [6, 7], and age-related brain changes [8].
Along with advancing age, impaired glymphatic function has been strongly linked to sleep disturbances, brain atrophy, and accelerated cognitive impairments. Notably, the bidirectional relationship between sleep quality, glymphatic function, and cognition is increasingly considered as a modifiable cycle that has great potential in the field of mechanism-based gerotherapeutics. For example, animal studies found that modulating glymphatic function through multisensory gamma stimulation or aerobic excise could promote the clearance of amyloid-beta, alleviate circadian disruption, and enhance cognitive performance in Alzheimer’s mice [9–13]. A few human pilot studies reported that repeated theta burst transcranial magnetic stimulation (TMS) over the left parietal cortex may enhance the glymphatic function and cognition in patients with mild cognitive impairment (MCI) [14]. Within this cycle, as a starting point, the association between chronological age and glymphatic function has been thoroughly researched. An age-related reduction in glymphatic function has been widely documented, with indirect measurement relying on the Diffusion Tensor Imaging Along the Perivascular Space (DTI-ALPS) index [15, 16]. It should be noted that brain morphometry [17–19] and the DTI-ALPS index [20, 21] present lateral differences in age-specific populations and correlate with specific cognitive domains. Beyond chronological age, brain age model has been developed to estimate an individual’s chronological age from imaging data [22–24]. This model contains two key measures: estimated brain age and the brain predicted age difference (brain-PAD). Brain-PAD is calculated by subtracting brain age from chronological age. The brain-PAD score can be interpreted in three ways: (1) a brain-PAD of zero reflects normal brain ageing (brain age = chronological age); (2) a negative brain-PAD indicates decelerated brain ageing (brain age < chronological age); and (3) a positive brain-PAD indicates accelerated brain ageing (brain age > chronological age) [25].
To date, however, glymphatic function and sleep quality in relation to different brain ageing phenotypes has received limited attention, and the potential modulatory role of brain age in the sleep–glymphatic function–cognition cycle remains understudied. Thus, this study sought to characterize brain age-specific differences in bilateral DTI-ALPS indices and to investigate their relationships with sleep features and cognitive functions in a well-characterized cohort of cognitively unimpaired adults.
Methods
Participants
All neuroimaging scans, cognitive performance, and brain health measures reported in this study were drawn from the Cambridge Centre for Ageing and Neuroscience (Cam-CAN) project (https://www.cam-can.org/). The Cam-CAN project, launched in October 2010, is designed to enhance our understanding of human ageing and study the optimal preservation of cognitive functions in late adulthood [26, 27]. For a comprehensive overview of the Cam-CAN project (completed in 2014), please see Taylor et al. [26].
In this retrospective analysis, we examined a subset of participants from Stage 2 of the population-based Cam-CAN cohort. The sample comprised 643 adults aged 18–87 years. To ensure a homogeneous group with preserved brain health, all included participants were right-handed native English speakers, scored ≥ 25 on the Mini-Mental State Examination (MMSE), and had no history of major neurological or psychiatric disorders. Among the initial participants, 21 cases were excluded because of the evident enlargement of ventricles and failures in automatically extracting fiber tracks during the processing of diffusion and structural magnetic resonance imaging (MRI) scans. Consequently, 582 participants successfully completed both structural and diffusion MRI scanning, along with assessments of lifestyle activities, subjective sleep quality, anxiety, depression, self-reported health conditions, and computerized reaction time. The final analytical cohort comprised 582 participants (292 female, 290 male), which was used to examine age-related changes in DTI-ALPS indices and to explore their associations with brain health-related measures.
Ethical Approval
This study was carried out in accordance with the recommendations of the university’s institutional review board (IRB). Ethical approval for all aspects of data collection and experimentation was obtained fromthe Cambridgeshire 2 Research Ethics Committee (East of England–Cambridge Central). The university’s IRB also provided explicit approval for open sharing of the anonymized data. All participants provided written informed consent prior to enrolment. Participants were informed that de-identified data would be used for publication purposes. The study was conducted in accordance with the ethical standards laid down in the Declaration of Helsinki (1964) and its subsequent amendments. Ethical approval for all the types of data collection and experiments was obtained from the Cambridgeshire 2 Research Ethics Committee (East of England-Cambridge Central) (REC reference number 10/H0308/50).
Collection of Clinical Data and MRI Images
Demographic data and brain health-related assessment scores were obtained via in-person interviews and comprehensive self-report questionnaires, using the standardized protocols established in the Stage 1 of the Cam-CAN project. The neuroimaging data analyzed in this study comprised structural T1-weighted MRI and diffusion tensor imaging (DTI) scans, which were retrieved from the Stage 2 repository of the Cam-CAN project. All MRI data were acquired on a 3.0-Tesla Siemens TIM Trio scanner equipped with a 32-channel head coil at the MRC Cognition and Brain Sciences Unit, University of Cambridge. The details of acquisition parameters can be found in the supplementary materials.
For structural analysis, T1-weighted scans were first registered to the Montreal Neurological Institute (MNI) space, with subsequent normalization performed as part of the DTI processing pipeline. This enabled the derivation of second-order diffusion metrics, including the full diffusion tensor matrix and fractional anisotropy (FA) values [28, 29].
Preprocessing of DTI Data and DTI-ALPS Index Calculation
DTI-ALPS was computed in MRtrix3 using custom scripts adapted from Taoka et al. [28]. The metric was derived from the mean diffusivity of specific regions located at the level of lateral ventricle body. As shown in Fig. 1, the calculations of left and right DTI-ALPS indices were based on the mean diffusivity of specific regions at the level of the lateral ventricle body. The DTI-ALPS index is defined using the following formula: DTI-ALPS index = mean (Dxproj, Dxassoc)/mean (Dyproj, Dzassoc). DTI-ALPS indices were computed independently for the left and right hemispheres, and a mean index was subsequently calculated as their average. The asymmetry index (AI), defined as AI = (Left − Right)/Global, was then derived to quantify interhemispheric lateralization [21].
Fig. 1.
Framework of structural and diffusion magnetic resonance imaging (MRI) analyses. On the basis of an individual’s MRI scans, the first step was to calculate brain age and brain predicted age difference (brain-PAD) based on quantified morphometric features. The second step was to calculate the bilateral Diffusion Tensor Imaging Along the Perivascular Space (DTI-ALPS) indices. The third step was to correlate brain health factors to brain age metrics and DTI-ALPS indices
MRI-Informed Brain Age Model
Estimated brain age was calculated using brainageR (v2.1), a software tool designed to generate a brain predicted age value for T1-weighted structural MRI scans (Available at: https://github.com/james-cole/brainageR). The brainageR workflow consists of two stages: preprocessing and prediction [30].
In the stage of preprocessing, raw T1-weighted MRI scans were segmented and normalized using the Statistical Parametric Mapping (SPM12) software (Available at: https://www.fil.ion.ucl.ac.uk/spm/software/spm12/) [30], which included bias field correction, segmentation into gray matter (GM), white matter (WM), and CSF, followed by registration to standard MNI152 space and smoothing with a 4-mm full width at half maximum (FWHM) kernel [31]. The normalized images were then loaded into R using the RNfiti package, where they were converted into vectors. The GM, WM, and CSF vectors were masked using a 0.3 threshold derived from the average image of the brainageR-specific template (n = 200, with 20 scans from each of 10 scanners) [32]. These masked vectors were subsequently combined for further analysis.
In the prediction stage, the brainageR model applied principal component analysis to the combined masked vectors to reduce dimensionality, retaining the top components that captured 80% of the variance, resulting in 435 components. These components were then used as inputs for a pretrained Gaussian Processes Regression (GPR) model with a radial basis function (RBF) kernel to estimate brain age. The GPR model was trained on 3377 healthy individuals aged 18–92 years (mean age 40.6 years, SD 21.4) and validated on 857 individuals for validation (r = 0.973, MAE 3.933 years) [32].
After validation, the trained brain age model was applied to the 582 participants in Cam-CAN dataset to estimate predicted brain age. For each individual, the brainageR model generated an estimated brain age, along with a 95% confidence interval (CI). The brain predicted age difference (brain-PAD) was then calculated by subtracting chronological age from the estimated brain age (i.e., brain age − chronological age).
Features of Subjective Sleep Quality
The Pittsburgh Sleep Quality Index (PSQI) is a 19-item self-rated instrument used to evaluate subjective sleep quality and the severity of sleep disturbances during the preceding month [33, 34]. The scale is composed of seven components, including: subjective sleep quality (C1), sleep latency (C2), sleep duration (C3), habitual sleep efficiency (C4), sleep disturbance (C5), use of sleeping medication (C6), and daytime dysfunction (C7). A higher PSQI total score reflects poorer subjective sleep quality and more pronounced sleep disturbances. Based on the components of PSQI, global sleep quality distinguishes good versus poor sleepers, sleep duration patterns (C3) distinguish long versus short sleepers, while sleep efficiency (C4) distinguishes high- versus low-efficiency sleepers [35]. A total score of PSQI ≤ 5 indicates a good sleeper with no clinically significant sleep difficulties, while a score > 5 classifies a poor sleeper [34]. Sleep efficiency, representing the percentage of time in bed actually spent asleep, was calculated using the following formula: Sleep efficiency = (Total sleep time/Time spent in bed) × 100%. High-efficiency sleepers: Sleep efficiency ≥ 85% (PSQI component 2 score = 0), and low-efficiency sleepers: Sleep efficiency < 85% (PSQI component 2 score ≥ 1).
Neuropsychiatric Symptoms
The Hospital Anxiety and Depression Scale (HADS) is a 14-item questionnaire used to assess the severity of both anxiety and depression concurrently, employing two distinct 7-item subscales [36]. Each item is rated on a 4-point Likert scale from 0 (absence of symptoms) to 4 (severe symptomatology), resulting in a composite score ranging from 0 to 56. The total score is interpreted as follows: below 17 indicates mild severity; 18–24, mild to moderate; and 25–30, moderate to severe.
General Health-Related Measures
Healthy data was collected through in-person interviews, which included physical measurements of weight, height, and body mass index (BMI), as well as assessments of systolic blood pressure (SBP), diastolic blood pressure (DBP), pulse pressure (PP), and pulse rate. Smoking and alcohol use were obtained via self-report questionnaires.
Cognitive Functions
Fluid Intelligence
The Cattell Culture Fair Intelligence Test (CFIT) is a non-verbal assessment designed to measure fluid intelligence by minimizing the influence of language and cultural background [37, 38]. It uses abstract shapes and patterns in tasks such as series completion and metrics to gauge a person’s innate reasoning capacity. The CFIT comprises four non-verbal, multiple-choice subtests of abstract reasoning: series completion, odd-one-out, matrix completion, and topological judgment. These subtests are designed to assess distinct components of fluid intelligence, with a higher total score indicating better fluid intelligence (IQ).
Accuracy and Processing Speed
A computer-based test that contains 50 trials for collecting reaction time (RT) is used to assess processing speed. In this test, participants were requested to view an image of a hand with blank circles above each finger, while resting their right hand on a response box with four buttons, one for each finger. When the index finger circle turns to black on the image, they should press with their index finger as quickly as possible. In this study, two measures, including accuracy and median RT, are used to evaluate the performance of processing speed [39]. RT is the completion time in milliseconds for a given trial recorded from stimulus onset to button press.
Intraindividual Variability (IIV)
Based on the RT collected during the computer-based test, IIV of RT indicates the short-term (i.e., trial-to-trial) fluctuations or variation in an individual’s RT performance [40]. In ageing studies, IIV of RT serves as a sensitive behavioral marker of the integrity of frontal-executive network and attention network [41]. The intraindividual coefficient of variation of reaction time (ICV-RT) is a common measure for assessing IIV while accounting for age-related slowing in processing speed. It is calculated using the following formula: ICV-RT = (SD of processing speed/mean of RT) × 100. Higher ICV-RT indicates greater short-term fluctuations or variation across the trials [42].
Statistical Analysis
The normality of variable distributions was evaluated using histograms, while equality of variances was tested with Levene’s procedure. Mean differences between different sleepers were then compared using either a two-sample t test or, when assumptions were not met, the Mann–Whitney U test. Pearson’s correlation was applied to analyze the relationships between DTI-ALPS indices, features of subjective sleep quality and mental health, and cognitive functions in older adults. Linear regression analyses were used to examine the associations between DTI-ALPS indices, subjective sleep quality, and cognitive functions, adjusting for demographic variables.
Mediation analyses were used to elucidate the mechanisms accounting for the observed relationships between DTI-ALPS indices and sleep quality features, while testing the mediating effects of brain age and chronological age on the pathway. A direct association between two variables can be represented as x → y. In mediation analysis, an intermediate variable (m) is introduced, forming an indirect pathway: x → m → y. Here, x serves as the independent variable, m as the mediator, and y as the dependent variable. On the basis of the results of correlation analysis, we hypothesize the impact of sleep quality (x) on cognitive functions is mediated by cognitive reserve or brain reserve. To quantitatively evaluate mediation, two linear regression equations are used. The slope coefficients from these equations, commonly referred to as path coefficients, indicate the magnitude and direction of the associations along each path. In this study, direct and indirect effects were calculated using the PROCESS for SPSS (version 5.0) with bootstrapping (5000 samples) confidence intervals. All statistical tests were two-tailed, and p values were corrected for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) procedure where appropriate. Statistical analyses were conducted using IBM SPSS Statistics 27.0, and results were visualized in Python.
Results
Participants
In total, the Cam-CAN dataset contained 582 participants, who were on average 53.43 ± 18.17 years old and 292 (50.2%) were female. The mean total score of PSQI was 5.28 ± 3.64, and 301 (51.7%) with good subjective sleep quality, 281 (48.3%) with poor sleep quality. Table 1 demonstrated a detailed descriptive summary of demographics, brain morphometry, DTI-ALPS indices, sleep quality features, health-related measures, and cognitive functions, stratified according to the statuses of subjective sleep quality. Compared to poor sleepers, good sleepers had significant higher right DTI-ALPS index (t = 2.199, p = 0.028), younger brain age (t = − 1.963, p = 0.049), thinner cortical thickness (t = 2.093, p = 0.038), better mental health (Anxiety: t = − 7.446, p < 0.001; Depression: t = − 6.612, p < 0.001), and higher fluid intelligence (t = 2.696, p = 0.007).
Table 1.
Demographics, MRI-informed brain features, and health-related characteristics
| Good sleepers (N = 301) | Poor sleepers (N = 279) | t value | p value | Effect size Cohen’s d |
|
|---|---|---|---|---|---|
| Demographics | |||||
| Age | 52.23 ± 17.72 | 54.71 ± 18.59 | − 1.647 | 0.101 | 0.137 |
| Sex (F/M) | 141/160 | 149/130 | − 1.661 | 0.097 | 0.138 |
| Handedness | 74.16 ± 43.88 | 72.32 ± 38.89 | 0.242 | 0.868 | 0.012 |
| Brain morphometry | |||||
| Cortical thickness (mm) | 3.99 ± 0.39 | 3.93 ± 0.32 | 2.093 | 0.038 | 0.174 |
| GMV (× 103 cm3) | 608.11 ± 62.59 | 599.01 ± 66.34 | 1.701 | 0.091 | 0.141 |
| CSF volume (× 103 cm3) | 257.96 ± 60.31 | 261.08 ± 63.23 | − 0.609 | 0.543 | 0.051 |
| Brain age (years) | 53.06 ± 13.73 | 55.36 ± 14.62 | − 1.963 | 0.049 | 0.166 |
| Brain-PAD | 0.83 ± 6.18 | 0.65 ± 6.81 | 0.328 | 0.743 | 0.027 |
| Glymphatic function | |||||
| Left DTI-ALPS | 1.46 ± 0.15 | 1.46 ± 0.14 | 0.199 | 0.842 | 0.017 |
| Right DTI-ALPS | 1.44 ± 0.14 | 1.41 ± 0.13 | 2.199 | 0.028 | 0.183 |
| DTI-ALPS | 1.45 ± 0.14 | 1.43 ± 0.13 | 1.233 | 0.218 | 0.102 |
| AI of DTI-ALPS | 0.007 ± 0.033 | 0.015 ± 0.034 | − 2.856 | 0.004 | 0.237 |
| Health-related characteristics | |||||
| BMI | 25.21 ± 3.72 | 25.84 ± 5.01 | − 1.632 | 0.103 | 0.145 |
| SBP (mmHg) | 120.15 ± 16.51 | 118.42 ± 15.71 | 1.202 | 0.231 | 0.107 |
| DBP (mmHg) | 73.39 ± 10.25 | 72.72 ± 9.99 | 0.744 | 0.457 | 0.066 |
| Pulse pressure (mmHg) | 46.76 ± 12.39 | 45.71 ± 12.19 | 0.964 | 0.336 | 0.086 |
| Pulse | 64.42 ± 10.23 | 66.06 ± 9.39 | − 1.836 | 0.067 | 0.163 |
| Smoking | 2.39 ± 1.49 | 2.71 ± 1.64 | − 2.366 | 0.018 | 0.197 |
| Alcohol | 5.27 ± 1.45 | 5.02 ± 1.58 | 1.937 | 0.053 | 0.161 |
| Anxiety | 4.07 ± 2.68 | 6.06 ± 3.71 | − 7.446 | < 0.001 | 0.619 |
| Depression | 2.09 ± 1.98 | 3.41 ± 2.81 | − 6.612 | < 0.001 | 0.551 |
| PSQI | 2.68 ± 1.16 | 8.09 ± 3.31 | − 26.744 | < 0.001 | 2.223 |
| Cognitive functions | |||||
| Accuracy | 0.96 ± 0.09 | 0.97 ± 0.05 | − 1.385 | 0.167 | 0.119 |
| Mean RT (seconds) | 0.58 ± 0.14 | 0.59 ± 0.15 | − 1.194 | 0.233 | 0.103 |
| IIV of RT | 0.22 ± 0.08 | 0.22 ± 0.07 | − 1.331 | 0.184 | 0.115 |
| Fluid intelligence | 32.93 ± 6.48 | 31.44 ± 6.27 | 2.696 | 0.007 | 0.227 |
Data are raw scores and presented as mean ± SD
MRI magnetic resonance imaging, GMV gray matter volume, DTI-ALPS diffusion tensor imaging along the perivascular space, BMI body mass index, SBP systolic blood pressure, DBP diastolic blood pressure, PSQI Pittsburgh Sleep Quality Index, RT reaction time, IIV of RT individual variability of reaction time
Chronological Age-Specific Changes in DTI-ALPS Indices
When chronological age was used as a continuous variable, age-related changes in the DTI-ALPS indices showed a nonlinear trajectory, increasing in early adulthood and decreasing in late adulthood. When we compared the DTI-ALPS indices among age-specific groups, the old-old adults had the lowest scores of DTI-ALPS indices (Global: F = 6.26, p < 0.001, η2 = 0.031; left side: F = 5.85, p = 0.001, η2 = 0.029; right side: F = 4.94, p = 0.002, η2 = 0.025) (Supplementary Table 1).
Brain Age-Specific Changes in DTI-ALPS Indices
The mean score of brain-PAD was 0.76 ± 6.49, and 238 (40.9%) with positive score, 344 (59.1%) with negative score. Compared to the individuals with negative brain-PAD, the adults with positive brain-PAD score had significantly lower left DTI-ALPS index (t = − 2.31, p = 0.021, Cohen’s d = 0.191), longer sleep latency (t = 4.14, p < 0.001, Cohen’s d = 0.346), and lower sleep efficiency (t = − 4.11, p < 0.001, Cohen’s d = 0.344) (Table 2). As shown in Fig. 2, when data were stratified according to sleep quality features, high-efficiency sleepers showed higher right DTI-ALPS index (t = 1.992, p = 0.047) in the adults with positive brain-PAD score, but both good sleepers (t = 2.159, p = 0.032) and high-efficiency sleepers (t = 2.271, p = 0.024) showed higher right DTI-ALPS index in the adults with negative brain-PAD score (Table 3).
Table 2.
Sleep quality, glymphatic function, and cognition in adults with different statuses of brain age
| Variables | Status of brain age | Brain-PAD+ versus brain-PAD− | |||
|---|---|---|---|---|---|
| Total (n = 582) | Brain-PAD+ (n = 238) | Brain-PAD− (n = 344) | p value (two-tailed) | Effect size (Cohen’d) | |
| General health data | |||||
| BMI | 25.54 ± 4.39 | 25.93 ± 4.21 | 25.19 ± 4.53 | 0.061 | 0.167 |
| SBP (mmHg) | 119.31 ± 16.12 | 123.21 ± 18.01 | 115.97 ± 13.49 | < 0.001 | 0.461 |
| DBP (mmHg) | 73.04 ± 10.11 | 72.58 ± 10.39 | 73.44 ± 9.86 | 0.337 | 0.085 |
| PP (mmHg) | 46.25 ± 12.29 | 50.63 ± 13.42 | 42.53 ± 9.82 | < 0.001 | 0.698 |
| Smoking | 2.54 ± 1.57 | 2.62 ± 1.67 | 2.46 ± 1.45 | 0.232 | 0.101 |
| Alcohol | 5.15 ± 1.51 | 5.17 ± 1.43 | 5.13 ± 1.61 | 0.723 | 0.029 |
| Anxiety | 5.03 ± 3.36 | 5.53 ± 3.38 | 4.49 ± 3.25 | < 0.001 | 0.312 |
| Depression | 2.72 ± 2.49 | 2.75 ± 2.38 | 2.71 ± 2.61 | 0.829 | 0.018 |
| Sleep quality parameters | |||||
| PSQI total score | 5.28 ± 3.64 | 5.38 ± 3.77 | 5.19 ± 3.51 | 0.519 | 0.054 |
| Latency (hours) | 1.31 ± 1.18 | 1.52 ± 1.26 | 1.11 ± 1.07 | < 0.001 | 0.346 |
| Duration (hours) | 6.94 ± 1.07 | 6.88 ± 1.08 | 6.99 ± 1.05 | 0.207 | 0.105 |
| Efficiency (%) | 84.78 ± 12.59 | 82.56 ± 13.24 | 86.82 ± 11.61 | < 0.001 | 0.344 |
| Glymphatic function | |||||
| Left DTI-ALPS | 1.46 ± 0.15 | 1.44 ± 0.14 | 1.47 ± 0.14 | 0.021 | 0.191 |
| Right DTI-ALPS | 1.43 ± 0.13 | 1.42 ± 0.13 | 1.43 ± 0.14 | 0.332 | 0.081 |
| DTI-ALPS | 1.44 ± 0.13 | 1.43 ± 0.13 | 1.45 ± 0.13 | 0.078 | 0.146 |
| AI of DTI-ALPS | 0.022 ± 0.07 | 0.016 ± 0.07 | 0.028 ± 0.06 | 0.048 | 0.164 |
| Cognitive functions | |||||
| Accuracy | 0.96 ± 0.08 | 0.96 ± 0.09 | 0.97 ± 0.06 | 0.179 | 0.116 |
| Mean RT (seconds) | 0.59 ± 0.14 | 0.65 ± 0.15 | 0.54 ± 0.11 | < 0.001 | 0.836 |
| IIV of RT | 0.22 ± 0.08 | 0.25 ± 0.09 | 0.21 ± 0.07 | < 0.001 | 0.552 |
| Fluid intelligence | 32.22 ± 6.57 | 29.27 ± 6.62 | 34.94 ± 5.22 | < 0.001 | 0.956 |
| Brain morphometry | |||||
| Cortical thickness (mm) | 3.96 ± 0.35 | 3.91 ± 0.25 | 4.03 ± 0.43 | < 0.001 | 0.333 |
| GMV (× 106 cm3) | 0.61 ± 0.06 | 0.59 ± 0.06 | 0.62 ± 0.06 | < 0.001 | 0.541 |
| CSF (× 106 cm3) | 0.26 ± 0.06 | 0.28 ± 0.07 | 0.24 ± 0.05 | < 0.001 | 0.585 |
Data are raw scores and presented as mean ± SD
Brain-PAD brain predicted age difference, BMI body mass index, SBP systolic blood pressure, DBP diastolic blood pressure, PP pulse pressure, PSQI Pittsburgh Sleep Quality Index, DTI-ALPS diffusion tensor imaging analysis along the perivascular space, AI asymmetry index, IIV intraindividual variability, RT reaction time, GMV gray matter volume, WMV white matter volume, CSF cerebrospinal fluid
Fig. 2.
Differences of global and DTI-ALPS indices in different types of sleepers in brain age groups. Calculated based on diffusion MRI scans (a–c), poor sleepers had lower right DTI-ALPS index than good sleepers (f). No differences of DTI-ALPS indices were found between long sleepers and short sleepers (g–i). The adults with lower sleep efficiency had lower right DTI-ALPS index, independent of the status of brain aging (l)
Table 3.
Brain age-specific differences in glymphatic function and cognition in different types of sleepers
| Brain age/imaging markers | Sleep quality | Sleep duration | Sleep efficiency | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Good | Poor | t value | p value | Long | Short | t value | p value | High | Low | t value | p value | |
| Brain-PAD+ | ||||||||||||
| Right DTI-ALPS | 1.43 ± 0.14 | 1.41 ± 0.13 | 0.958 | 0.339 | 1.43 ± 0.14 | 1.39 ± 0.13 | 1.733 | 0.084 | 1.44 ± 0.14 | 1.40 ± 0.13 | 1.992 | 0.047 |
| Left DTI-ALPS | 1.44 ± 0.14 | 1.44 ± 0.15 | 0.313 | 0.754 | 1.45 ± 0.15 | 1.43 ± 0.14 | 1.025 | 0.306 | 1.46 ± 0.15 | 1.43 ± 0.14 | 1.649 | 0.101 |
| DTI-ALPS | 1.43 ± 0.13 | 1.42 ± 0.13 | 0.317 | 0.751 | 1.44 ± 0.13 | 1.41 ± 0.12 | 1.452 | 0.148 | 1.45 ± 0.13 | 1.42 ± 0.13 | 1.937 | 0.054 |
| Anxiety | 3.41 ± 2.53 | 5.66 ± 3.54 | − 6.129 | < 0.001 | 4.17 ± 2.99 | 5.24 ± 3.73 | − 2.532 | 0.012 | 4.21 ± 3.11 | 4.76 ± 3.41 | − 1.394 | 0.164 |
| Depression | 2.12 ± 1.92 | 3.43 ± 2.64 | − 4.736 | < 0.001 | 2.45 ± 2.18 | 3.29 ± 2.65 | − 2.733 | 0.007 | 2.51 ± 2.39 | 2.91 ± 2.32 | − 1.422 | 0.156 |
| Accuracy | 0.96 ± 0.11 | 0.95 ± 0.07 | 1.272 | 0.204 | 0.97 ± 0.11 | 0.96 ± 0.12 | 1.251 | 0.212 | 0.96 ± 0.11 | 0.95 ± 0.07 | 0.679 | 0.498 |
| Reaction time | 0.64 ± 0.15 | 0.65 ± 0.15 | − 0.833 | 0.406 | 0.63 ± 0.15 | 0.67 ± 0.15 | − 1.738 | 0.083 | 0.62 ± 0.15 | 0.67 ± 0.14 | − 2.444 | 0.015 |
| IIV-RT | 0.24 ± 0.09 | 0.25 ± 0.08 | − 0.672 | 0.502 | 0.24 ± 0.09 | 0.26 ± 0.08 | − 1.161 | 0.247 | 0.24 ± 0.09 | 0.25 ± 0.08 | − 1.351 | 0.178 |
| Intelligence | 29.89 ± 6.37 | 28.56 ± 6.47 | 1.652 | 0.099 | 29.85 ± 6.65 | 28.38 ± 6.18 | 1.704 | 0.091 | 30.03 ± 6.55 | 28.75 ± 0.11 | 1.611 | 0.109 |
| Brain-PAD− | ||||||||||||
| Right DTI-ALPS | 1.45 ± 0.13 | 1.41 ± 0.13 | 2.159 | 0.032 | 1.44 ± 0.14 | 1.42 ± 0.13 | 1.121 | 0.263 | 1.44 ± 0.13 | 1.41 ± 0.12 | 2.271 | 0.024 |
| Left DTI-ALPS | 1.48 ± 0.15 | 1.47 ± 0.14 | 0.597 | 0.551 | 1.48 ± 0.15 | 1.46 ± 0.13 | 0.643 | 0.521 | 1.48 ± 0.15 | 1.46 ± 0.14 | 1.104 | 0.257 |
| DTI-ALPS | 1.46 ± 0.14 | 1.44 ± 0.12 | 0.417 | 0.154 | 1.46 ± 0.13 | 1.44 ± 0.12 | 0.929 | 0.354 | 1.46 ± 0.13 | 1.43 ± 0.12 | 1.769 | 0.078 |
| Anxiety | 4.69 ± 2.67 | 6.43 ± 3.81 | − 4.606 | < 0.001 | 5.22 ± 2.97 | 6.49 ± 4.34 | − 2.765 | 0.006 | 5.15 ± 2.91 | 6.27 ± 4.09 | − 2.732 | 0.007 |
| Depression | 2.05 ± 2.04 | 3.41 ± 2.94 | − 4.629 | < 0.001 | 2.49 ± 2.39 | 3.38 ± 3.14 | − 2.539 | 0.012 | 2.33 ± 2.39 | 3.45 ± 2.87 | − 3.537 | < 0.001 |
| Accuracy | 0.98 ± 0.03 | 0.97 ± 0.09 | 0.591 | 0.555 | 0.97 ± 0.07 | 0.96 ± 0.08 | 0.454 | 0.651 | 0.97 ± 0.08 | 0.96 ± 0.03 | 0.658 | 0.511 |
| Reaction time | 0.53 ± 0.09 | 0.54 ± 0.11 | − 1.273 | 0.204 | 0.53 ± 0.11 | 0.54 ± 0.09 | − 0.752 | 0.453 | 0.53 ± 0.09 | 0.55 ± 0.12 | − 1.659 | 0.098 |
| IIV-RT | 0.19 ± 0.06 | 0.21 ± 0.07 | − 1.511 | 0.132 | 0.21 ± 0.06 | 0.21 ± 0.07 | − 0.679 | 0.498 | 0.19 ± 0.06 | 0.21 ± 0.07 | − 1.215 | 0.225 |
| Intelligence | 35.75 ± 4.74 | 34.09 ± 5.59 | 2.739 | 0.007 | 35.25 ± 5.04 | 34.06 ± 5.73 | 1.657 | 0.099 | 35.57 ± 4.72 | 33.72 ± 6.01 | 2.855 | 0.005 |
Data are raw scores and presented as mean ± SD
Brain-PAD brain predicted age difference, DTI-ALPS diffusion tensor imaging analysis along the perivascular space, IIV of RT intraindividual variability of reaction time
Brain Age and Sleep–Glymphatic function–Cognition Cycle
In general, estimated brain age was markedly correlated with lower DTI-ALPS indices (Left: r = − 0.091, p = 0.028; Right: r = − 0.09, p = 0.029; Global: r = − 0.097, p = 0.019). Increased brain-PAD score was correlated with lower left DTI-ALPS index (r = − 0.084, p = 0.042). Within the adults with positive brain-PAD score, right DTI-ALPS index was correlated with subjective sleep quality parameters and mental health-related features (Fig. 3a). No significant correlations were observed between DTI-ALPS indices and cognitive functions. Within the adults with negative brain-PAD score, global and right DTI-ALPS indices were significantly correlated with cognitive functions, including fluid intelligence, reaction time, and IIV of RT (Fig. 3b). No significant correlations were observed between DTI-ALPS indices and subjective sleep quality features.
Fig. 3.
Brain age-specific effects on the sleep–glymphatic function–cognition associations. a Adults with a positive brain-PAD score displayed a robust correlation between DTI-ALPS indices and sleep quality features, whereas b adults with a negative brain-PAD score showed a reliable correlation between DTI-ALPS indices and cognitive functions. The color scale depicts the strength of the Pearson correlation coefficient. Asterisk depict the statistical significance of the correlation: *p < 0.05; **p < 0.005. Brain-PAD brain predicted age difference, DTI-ALPS diffusion tensor imaging along the perivascular space
Analyses of Interaction and Mediation Effects
According to univariate analyses of variance, a small but significant main effect of brain age was found on the left DTI-ALPS index (partial η2 range 0.007–0.009; observed power range 0.529–0.638). Significant main effect of sleep statuses was consistently found on the right DTI-ALPS index (sleep efficiency: partial η2 = 0.016; observed power = 0.853; sleep duration: partial η2 = 0.007; observed power = 0.521; sleep quality: partial η2 = 0.008; observed power = 0.584). No interaction effect between brain age and sleep status was found on the DTI-ALPS indices.
Focusing on DTI-ALPS indices and sleep quality parameters, we conducted mediation analyses to estimate the three paths: a (association between DTI-ALPS indices and brain age or chronological age), b (association between brain age or chronological age and sleep quality parameters), and c (association between DTI-ALPS indices and sleep quality parameters), as well as the indirect effect (a to b) among the overall associations. As shown in Fig. 4, the associations between left DTI-ALPS index and sleep efficiency were completely mediated by brain age (direct effect: β = 5.93, p = 0.089, 95% CI − 0.902, 12.761; indirect effect: β = 1.904, 95% CI 0.086, 4.031) and chronological age (direct effect: β = 5.655, p = 0.104, 95% CI − 1.17, 12.481; indirect effect: β = 2.179, 95% CI 0.355, 4.337). The associations between global DTI-ALPS index and sleep efficiency were partially mediated by brain age (direct effect: β = 8.71, p = 0.024, 95% CI 1.166, 16.244; indirect effect: β = 2.226, 95% CI 0.267, 4.512) and chronological age (direct effect: β = 8.55, p = 0.026, 95% CI 1.021, 16.076; indirect effect: β = 2.383, 95% CI 0.368, 4.727) (Fig. 4). The associations between right DTI-ALPS index and sleep efficiency were only partially mediated by brain age (direct effect: β = 9.92, p = 0.008, 95% CI 2.616, 17.217; indirect effect: β = 1.996, 95% CI 0.129, 4.185), not chronological age. No mediation effect of brain age or chronological age was found on the associations between DTI-ALPS indices and PSQI total score (Supplementary Fig. 1).
Fig. 4.
Mediation analyses examining the relationships among age, DTI-ALPS indices, and sleep efficiency. Panels a–c show mediation models with estimated brain age as the mediator, while panels d–f show models with chronological age as the mediator. Path coefficients (β) and p values were reported for each relationship. DTI-ALPS diffusion tensor imaging along the perivascular space
Predictive Performance of DTI-ALPS Indices
In the explorative regression analyses, global and bilateral DTI-ALPS indices were included as predictors. As shown in Table 4, global and right DTI-ALPS indices were significant predictors for sleep quality statuses in adults with negative brain-PAD, not in adults with positive brain-PAD.
Table 4.
Estimated effect sizes for the DTI-ALPS indices included in the predictive models of sleep quality parameters
| Brain age/glymphatic function | Sleep quality | Sleep duration | Sleep efficiency | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| B | 95% CI | p value | B | 95% CI | p value | B | 95% CI | p value | ||||
| Lower | Upper | Lower | Upper | Lower | Upper | |||||||
| All participants | ||||||||||||
| DTI-ALPS | − 2.865 | − 5.112 | − 0.618 | 0.013 | 0.925 | 0.267 | 1.583 | 0.006 | 10.345 | 2.531 | 18.161 | 0.010 |
| Left DTI-ALPS | − 1.713 | − 3.752 | 0.326 | 0.099 | 0.702 | 0.104 | 1.299 | 0.021 | 7.256 | 0.173 | 14.339 | 0.045 |
| Right DTI-ALPS | − 3.509 | − 5.684 | − 1.335 | 0.002 | 0.952 | 0.315 | 1.591 | 0.003 | 11.481 | 3.902 | 19.058 | 0.003 |
| Brain-PAD+ | ||||||||||||
| DTI-ALPS | − 1.635 | − 5.001 | 1.731 | 0.341 | 0.789 | − 0.175 | 1.753 | 0.108 | 9.153 | − 2.801 | 21.106 | 0.133 |
| Left DTI-ALPS | − 1.068 | − 4.135 | 2.001 | 0.494 | 0.705 | − 0.176 | 1.587 | 0.116 | 6.411 | − 4.485 | 17.305 | 0.248 |
| Right DTI-ALPS | − 1.986 | − 5.211 | 1.238 | 0.266 | 0.705 | − 0.218 | 1.627 | 0.134 | 10.381 | − 1.065 | 21.827 | 0.075 |
| Brain-PAD− | ||||||||||||
| DTI-ALPS | − 3.961 | − 6.994 | − 0.926 | 0.011 | 1.002 | 0.092 | 1.911 | 0.031 | 9.081 | − 0.986 | 19.148 | 0.077 |
| Left DTI-ALPS | − 2.221 | − 4.977 | 0.535 | 0.114 | 0.639 | − 0.184 | 1.463 | 0.128 | 5.341 | − 3.767 | 14.451 | 0.249 |
| Right DTI-ALPS | − 4.975 | − 7.924 | − 2.026 | 0.001 | 1.167 | 0.279 | 2.055 | 0.010 | 11.118 | 1.291 | 20.946 | 0.027 |
Data are raw scores and presented as mean ± SD
DTI-ALPS diffusion tensor imaging along the perivascular space, Brain-PAD brain predicted age difference, CI confidence interval
Discussion
In this study, we systematically investigated the brain age metrics and their effects on the associations between subjective sleep quality, glymphatic function, and cognition in a well-characterized community-dwelling cohort. Through quantifying MRI-informed brain age metrics and DTI-ALPS indices, four major findings observed in this study provide new evidence and clinical insights for this field. First, the lateral differences in DTI-ALPS indices decreased with chronological age and brain age. Second, adults with accelerated ageing (i.e., brain-PAD+) demonstrated lower left DTI-ALPS index independent of sex. Third, the robust associations between DTI-ALPS indices and sleep efficiency were rooted in the right hemisphere. Fourth, brain age exerted specific effects on the associations between sleep quality, glymphatic function, and cognition: adults with a positive brain-PAD score exhibited a robust correlation between DTI-ALPS index and sleep quality features, whereas those with a negative brain-PAD score showed a reliable correlation between DTI-ALPS index and cognition, highlighting a selective effect confined to brain ageing.
Chronological Age, Brain Age, and DTI-ALPS Indices
Throughout adulthood, the DTI-ALPS indices followed an inverted U-shaped trajectory, increasing gradually until approximately 50 years of chronological age and declining thereafter. Chronological age-related DTI-ALPS changes have been thoroughly investigated in normal ageing populations. Although the global DTI-ALPS index maintained a mean value of approximately 1.45, leftward asymmetry remained consistent across all age-specific groups. A lateralized DTI-ALPS index has been found to correlate with ageing brain health. For instance, left dominance in DTI-ALPS index is common in healthy ageing adults [20] and Parkinson’s disease [21], which links to handedness, motor function, and language. In addition to lateralized function, the hemispheric asymmetry reduction in older adults (HAROLD) model offers another neuroscientific explanation of age-related changes in structural and functional hemispheric asymmetry, positing a compensatory mechanism in the ageing brain [43]. Although no significant differences were detected in the AI of DTI-ALPS index, a clear trend toward reduced hemispheric asymmetry was observed. Interestingly, leftward asymmetry in DTI-ALPS index was also detected across adults with different statues of brain ageing, although it was reduced in those with a negative brain-PAD score. In parallel, adults with accelerated ageing (brain-PAD+) showed a significant lower left DTI-ALPS index.
The brain-PAD score, defined as the difference between estimated brain age and chronological age, captures cumulative genetic, vascular, and lifestyle exposures, and indicates risk for poor brain health [44], progression from normal ageing to MCI [24], and multiple neurological and psychiatric conditions [30, 45–47]. A positive brain-PAD score indicates accelerated brain ageing, while a negative brain-PAD score suggests relative resilience or delayed ageing [23]. Higher brain-PAD score correlated with lower left-hemisphere DTI-ALPS index, indicating that the global DTI-ALPS index may underestimate early changes in brain resilience and glymphatic function in older adults. These complex relationships between DTI-ALPS indices and brain age metrics warrant further investigation within the broader context of brain health.
DTI-ALPS Indices and Brain Health
The glymphatic system plays a fundamental role in clearing toxic proteins (e.g., beta-amyloid) and maintaining cerebral homeostasis, which works most effectively during sleep [48, 49]. Since its first publication in 2017, the DTI-ALPS index has been widely used in patients with neurological and cerebrovascular diseases. This index quantifies perivascular fluid diffusion, serving as a noninvasive, indirect measure of glymphatic clearance [28]. Within the sleep–glymphatic function cycle, the relationship between DTI-ALPS index and sleep quality has been well studied. Several large-scale studies have confirmed the link between lower DTI-ALPS index and poor sleep quality or insomnia [15, 50, 51]. However, no studies to date have examined the lateral DTI-ALPS differences in connection with sleep quality features and cognitive functions. Beyond global DTI-ALPS measurement, we observed, for the first time, significant asymmetry in the DTI-ALPS index among adults with poor sleep quality, particularly in relation to global sleep quality and shorter sleep duration. Compared to age- and sex-matched good sleepers, poor sleepers had a markedly lower right-sided DTI-ALPS index. Moreover, robust associations were found between right DTI-ALPS index and sleep efficiency across all participants.
Unlike chronological age, brain age metrics are closely related to sleep quality across multiple timescales. For example, several studies have reported that poor sleep quality accelerates brain ageing, as measured by brain-PAD score [52, 53]. Additionally, increased brain age has been observed in the healthy young adults following acute sleep deprivation, and this effect was found to be reversible with recovery sleep [54]. Increased chronological age is an unmodifiable risk factor for neurodegeneration; however, brain age may be modifiable and preserved through disease-modifying interventions or gerotherapeutics during the ageing process. Importantly, our analysis revealed, for the first time, two distinct patterns of sleep–glymphatic function–cognition associations based on brain age: adults with a positive brain-PAD score displayed correlations between DTI-ALPS indices and sleep quality features, whereas those with a negative brain-PAD score showed correlations between DTI-ALPS indices and cognitive functions.
In-depth investigation of the “sleep–glymphatic function–cognition” connections is crucial, as these relationships may inform the development of targeted interventions aimed at maintaining cognitive functions and managing comorbidities in preclinical populations. As highlighted in Life’s Essential 8 [55], sleep quality is now viewed as a core pillar of brain health and a modifiable lifestyle factor that contributes to accelerated brain ageing. Moreover, untreated poor sleep quality or sleep disturbances have been identified as key risk factors for cognitive decline and Alzheimer’s disease [56].
To further investigate the multifaceted relationships among brain age, glymphatic function, and key sleep quality features, we performed mediation analyses. Interestingly, brain age was found to partially mediate the relationship between sleep efficiency and both global DTI-ALPS and right DTI-ALPS indices, yet it completely mediated the relationship between sleep efficiency and left DTI-ALPS index. No mediation effect of brain age was observed in the relationship between DTI-ALPS indices and global sleep quality. Among the sleep quality features examined, only the link between left DTI-ALPS index and sleep efficiency was significantly explained by brain age. Through this mediation pathway, the left DTI-ALPS index may serve as a proxy for sleep efficiency, even in cognitive unimpaired adults. This study provides the first demonstration that brain age mediates the relationship between glymphatic function and sleep efficiency. These findings offer valuable clinical implications for targeting modifiable factors in the development of mechanism-based gerotherapeutics. Grounded in the coupling of brain age and sleep quality, individual brain age metrics may potentially serve both as an indicator and a valuable clinical outcome, thereby broadening our understanding of sleep–glymphatic function connections in older adults.
Although our study provides novel findings and gerotherapeutic insights into brain age-specific effects on the relationships among lateralized DTI-ALPS, sleep quality, and cognitive functions in normal ageing, several limitations should be acknowledged. First, the cross-sectional nature of the Cam-CAN dataset precludes causal inference and limits our ability to examine ageing-related effects on DTI-ALPS indices, sleep quality, and cognition over time. Future studies incorporating longitudinal assessments of DTI-ALPS indices and brain age metrics are therefore warranted to elucidate the dynamic link between glymphatic function and cognitive decline. Second, the DTI-ALPS index is calculated based on the water diffusion in white matter regions adjacent to the lateral ventricle, where the orientations of fibers may differ between the hemispheres. Although no differences were found in the FA values across brain age groups and sleep groups, the nature of DTI-ALPS could indirectly measure the glymphatic function that may be influenced by CSF and white matter microstructure [57, 58]. Third, major potential confounders, such as genetic factors (e.g., apolipoprotein E, TREM2), were disregarded, as our current aim was not to train a prediction model in a subgroup defined by specific allele status. Fourth, sleep quality in this study was assessed using questionaries, which cannot capture objective circadian rhythms or chronotypes. Fifth, a single modality of MRI data was used in the calculation of the brain age model. While recent multimodal brain age models integrating structural, diffusion, and functional MRI have demonstrated improved performance, the present study utilized T1-weighted structural imaging owing to its widespread availability and suitability for clinical cohorts. Finally, we did not apply bias correction to brain age, which may lead to regression-to-the-mean effects on brain-PAD scores, particularly at the extremes of the age distribution. Future studies should incorporate correction methods and consider potential nonlinear relationships between brain age and DTI-ALPS to enhance the accuracy of brain age predictions in ageing cohorts.
Conclusions
This study advances the applications of structural and diffusion imaging toward the assessments of glymphatic function and brain age metrics. Our findings suggest that the lateralized DTI-ALPS index holds considerable promise as an indirect measure of glymphatic clearance, with potential utility for predicting brain health, sleep efficiency, and cognition at individual level. Our analyses further revealed two distinct profiles of sleep–glymphatic function–cognition associations based on brain ageing status. Most notably, preserved brain age emerged as a protective factor against age-related declines in glymphatic function and sleep efficiency, highlighting it as a promising therapeutic target for enhancing brain resilience.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
The authors would like to thank all the participants, the principal investigators, and the whole research team of the Cambridge Centre for Ageing and Neuroscience (Cam-CAN) project. We are especially grateful to the participants for their participation and contribution. The Cam-CAN funding was provided by the UK Biotechnology and Biological Sciences Research Council (Grant number BB/H008217/1), together with the support from the UK Medical Research Council and University of Cambridge, UK.
Author Contribution
Hanna Lu, Zeyan Li, and Xi Ni conceptualized this project. Zeyan Li and Hanna Lu conducted the neuroimaging data processing. Xi Ni performed the statistical analysis. Hanna Lu prepared the tables, figures, and wrote the original draft of this manuscript. Liwei Guo, Ben Chen, Xiaomei Zhong, Yanling Zhou, and Yuping Ning reviewed and edited the manuscript. Hanna Lu, Zeyan Li, Liwei Guo, Xi Ni, Ben Chen, Xiaomei Zhong, Yanling Zhou, and Yuping Ning read and approved the final manuscript.
Funding
This work was supported by the Healthy Longevity Catalyst Awards (Hong Kong) (Project Number HLCA/M-407/23) and Research Data Management Development Fund, The Chinese University of Hong Kong (Project Number 4730346). The Rapid Service Fee was funded by the authors. The corresponding author had full access to all the data in the study and takes final responsibility for the decision to submit the manuscript for publication. The funders had no role in the study design, data collection, data analysis, or preparation of this article.
Data Availability
The raw data can be requested and downloaded via the official website of Cam-CAN project (https://www.cam-can.org/). The analyzed neuroimaging data that supports the findings of this study are available on request from the corresponding author.
Declarations
Conflict of Interest
Hanna Lu, Zeyan Li, Liwei Guo, Xi Ni, Ben Chen, Xiaomei Zhong, Yanling Zhou, and Yuping Ning declare no competing interests.
Ethical Approval
This study was carried out in accordance with the recommendations of the university’s institutional review board (IRB). Ethical approval for all aspects of data collection and experimentation was obtained from the Cambridgeshire 2 Research Ethics Committee (East of England–Cambridge Central). The university’s IRB also provided explicit approval for open sharing of the anonymized data. All participants provided written informed consent prior to enrolment. Participants were informed that de-identified data would be used for publication purposes. The study was conducted in accordance with the ethical standards laid down in the Declaration of Helsinki (1964)and its subsequent amendments. Ethical approval for all the types of data collection and experiments was obtained from the Cambridgeshire 2 Research Ethics Committee (East of England-Cambridge Central) (REC reference number 10/H0308/50).
Footnotes
Prior Presentation: Preliminary results, including the associations between cognition and sleep quality features, have been presented at BRAIN & BRAIN PET 2025 in Seoul, South Korea (June 1–4, 2025). However, the comprehensive analyses and results reported in this study have not been previously presented or published.
References
- 1.Plog BA, Nedergaard M. The glymphatic system in central nervous system health and disease: past, present, and future. Annu Rev Pathol Mech Dis. 2018;13:379–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Chong PL, Garic D, Shen MD, Lundgaard I, Schwichtenberg AJ. Sleep, cerebrospinal fluid, and the glymphatic system: a systematic review. Sleep Med Rev. 2022;61:101572. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Siow TY, Toh CH, Hsu JL, et al. Association of sleep, neuropsychological performance, and gray matter volume with glymphatic function in community-dwelling older adults. Neurology. 2022;98(8):e829–38. [DOI] [PubMed] [Google Scholar]
- 4.Silva I, Silva J, Ferreira R, Trigo D. Glymphatic system, AQP4, and their implications in Alzheimer’s disease. Neurol Res Pract. 2021;3(1):5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Plog BA, Lou N, Pierre CA, et al. When the air hits your brain: decreased arterial pulsatility after craniectomy leading to impaired glymphatic flow. J Neurosurg. 2019;133(1):210–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Mendelsohn AR, Larrick JW. Sleep facilitates clearance of metabolites from the brain: glymphatic function in aging and neurodegenerative diseases. Rejuvenation Res. 2013;16(6):518–23. [DOI] [PubMed] [Google Scholar]
- 7.Aquino G, Palagini L, Alfì G, et al. The interplay between the sleep slow oscillation and cerebrospinal fluid: new vistas for insomnia research. J Sleep Res. 2025;34(5):e70069. [DOI] [PubMed] [Google Scholar]
- 8.Yin F, Zeng B, Nie L, et al. Aging as a key determinant of glymphatic system function in older adults. J Gerontol A Biol Sci Med Sci. 2025;80(11):glaf176. [DOI] [PubMed] [Google Scholar]
- 9.Murdock MH, Yang CY, Sun N, et al. Multisensory gamma stimulation promotes glymphatic clearance of amyloid. Nature. 2024;627(8002):149–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Wu W, Zhao Y, Cheng X, et al. Modulation of glymphatic system by visual circuit activation alleviates memory impairment and apathy in a mouse model of Alzheimer’s disease. Nat Commun. 2025;16(1):63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Liang S, Liu H, Wang X, et al. Aerobic exercise improves clearance of amyloid-β via the glymphatic system in a mouse model of Alzheimer’s disease. Brain Res Bull. 2025;222:111263. [DOI] [PubMed] [Google Scholar]
- 12.Hu Y, Niu L, Chen Y, et al. Voluntary wheel running exercise improves sleep disorder, circadian rhythm disturbance, and neuropathology in an animal model of Alzheimer’s disease. Alzheimers Dement. 2025;21(6):e70314. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Mee-Inta O, Chiang YY, Tsai SF, et al. Running exercise mitigates amyloidosis in 5xFAD mice by improving the structure and function of the meningeal lymphatic system. Acta Neuropathol Commun. 2026;14(1):126. [DOI] [PMC free article] [PubMed]
- 14.Sundman MH, Liu Y, Chen NK, Chou YH. The glymphatic system as a therapeutic target: TMS-induced modulation in older adults. Front Aging Neurosci. 2025;17:1597311. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Clark O, Delgado-Sanchez A, Cullell N, Correa SA, Krupinski J, Ray N. Diffusion tensor imaging analysis along the perivascular space in the UK biobank. Sleep Med. 2024;119:399–405. [DOI] [PubMed] [Google Scholar]
- 16.Satpathi S, Reid RI, Przybelski SA, et al. Evaluation and interpretation of DTI-ALPS, a proposed surrogate marker for glymphatic clearance, in a large population-based sample. Alzheimers Res Ther. 2025;17(1):191. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Witelson SF, Beresh H, Kigar DL. Intelligence and brain size in 100 postmortem brains: sex, lateralization and age factors. Brain. 2006;129(2):386–98. [DOI] [PubMed] [Google Scholar]
- 18.Olulade OA, Seydell-Greenwald A, Chambers CE, et al. The neural basis of language development: changes in lateralization over age. Proc Natl Acad Sci USA. 2020;117(38):23477–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Hennessee JP, Webb CE, Chen X, Kennedy KM, Wig GS, Park DC. Relationship of prefrontal brain lateralization to optimal cognitive function differs with age. Neuroimage. 2022;264:119736. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Perlaki G, Dudás B, Horváth R, et al. Diffusion along the perivascular space influenced by handedness and language lateralisation. Brain Commun. 2025;7(4):fcaf252. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Li Z, Miao X, Zhang Q, et al. Diffusion-tensor MRI study of the relationship between glymphatic system asymmetry and onset lateralization in Parkinson’s disease. NPJ Parkinsons Dis. 2025;11(1):218. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Franke K, Ziegler G, Klöppel S, Gaser C, Alzheimer’s Disease Neuroimaging Initiative. Estimating the age of healthy subjects from T1-weighted MRI scans using kernel methods: exploring the influence of various parameters. Neuroimage. 2010;50(3):883–92. [DOI] [PubMed] [Google Scholar]
- 23.Cole JH, Franke K. Predicting age using neuroimaging: innovative brain ageing biomarkers. Trends Neurosci. 2017;40(12):681–90. [DOI] [PubMed] [Google Scholar]
- 24.Lu H, Li J. MRI-informed machine learning-driven brain age models for classifying mild cognitive impairment converters. J Cent Nerv Syst Dis. 2024;16:11795735241266556. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Lu H, Li J, Chan SSM, et al. Predictive values of pre-treatment brain age models to rTMS effects in neurocognitive disorder with depression: secondary analysis of a randomised sham-controlled clinical trial. Dialogues Clin Neurosci. 2024;26(1):38–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Taylor JR, Williams N, Cusack R, et al. The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) data repository: structural and functional MRI, MEG, and cognitive data from a cross-sectional adult lifespan sample. Neuroimage. 2017;144:262–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Henson RN. Fifteen years on: a review of the Cam-CAN study of the cognitive neuroscience of ageing. GeroScience. 2026. 10.1007/s11357-026-02096-z. [DOI] [PubMed] [Google Scholar]
- 28.Taoka T, Masutani Y, Kawai H, et al. Evaluation of glymphatic system activity with the diffusion MR technique: diffusion tensor image analysis along the perivascular space (DTI-ALPS) in Alzheimer’s disease cases. Jpn J Radiol. 2017;35(4):172–8. [DOI] [PubMed] [Google Scholar]
- 29.Taoka T, Ito R, Nakamichi R, Nakane T, Kawai H, Naganawa S. Diffusion tensor image analysis along the perivascular space (DTI-ALPS): revisiting the meaning and significance of the method. Magn Reson Med Sci. 2024;23(3):268–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Biondo F, Jewell A, Pritchard M, et al. Brain-age is associated with progression to dementia in memory clinic patients. NeuroImage Clin. 2022;36:103175. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Ashburner J, Barnes G, Chen CC, et al. SPM12 manual. London: Wellcome Trust Centre for Neuroimaging; 2014. p. 53. [Google Scholar]
- 32.Hobday H, Cole JH, Stanyard RA, et al. Tissue volume estimation and age prediction using rapid structural brain scans. Sci Rep. 2022;12(1):12005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Carpenter JS, Andrykowski MA. Psychometric evaluation of the Pittsburgh sleep quality index. J Psychosom Res. 1998;45(1):5–13. [DOI] [PubMed] [Google Scholar]
- 34.Ni X, Yuen YS, Li Z, et al. Focused low-intensity hippocampal transcranial ultrasound stimulation (TUS) for sleep disturbances in patients with chronic tinnitus: a study protocol for a pilot randomized controlled trial. PLoS ONE. 2025;20(8):e0329950. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Ni X, Yang NS, Yuen YS, et al. HK-ASAP study: protocol for a prospective cohort study on sleep quality, brain reserve and cognitive phenotypes in community-dwelling older adults. BMC Geriatr. 2026;26(1):625. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Snaith RP. The hospital anxiety and depression scale. Health Qual Life Outcomes. 2003;1(1):29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Johnson W, Nijenhuis J, Bouchard TJ. Still just 1 g: consistent results from five test batteries. Intelligence. 2008;36(1):81–95. [Google Scholar]
- 38.Mitchell DJ, Mousley A, Shafto MA, et al. Neural contributions to reduced fluid intelligence across the adult lifespan. J Neurosci. 2023;43(2):293–307. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Lu H, Chan SS, Fung AW, Lam LC. Beyond a differential diagnosis: cognitive and morphometric decoding of information processing speed in senior adults with DSM-5 mild neurocognitive disorders. J Alzheimers Dis. 2017;58(3):927–37. [DOI] [PubMed] [Google Scholar]
- 40.MacDonald SW, Hultsch DF, Dixon RA. Performance variability is related to change in cognition: evidence from the Victoria Longitudinal Study. Psychol Aging. 2003;18(3):510. [DOI] [PubMed] [Google Scholar]
- 41.Lu H, Fung AW, Chan SS, Lam LC. Disturbance of attention network functions in Chinese healthy older adults: an intra-individual perspective. Int Psychogeriatr. 2016;28(2):291–301. [DOI] [PubMed] [Google Scholar]
- 42.Lu H, Chan SS, Lam LC. ‘Two-level’ measurements of processing speed as cognitive markers in the differential diagnosis of DSM-5 mild neurocognitive disorders (NCD). Sci Rep. 2017;7(1):521. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Cabeza R. Hemispheric asymmetry reduction in older adults: the HAROLD model. Psychol Aging. 2002;17(1):85. [DOI] [PubMed] [Google Scholar]
- 44.Wagen AZ, Coath W, Keshavan A, et al. Life course, genetic, and neuropathological associations with brain age in the 1946 British Birth Cohort: a population-based study. Lancet Healthy Longev. 2022;3(9):e607–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Cole JH, Raffel J, Friede T, et al. Longitudinal assessment of multiple sclerosis with the brain age paradigm. Ann Neurol. 2020;88(1):93–105. [DOI] [PubMed] [Google Scholar]
- 46.Kim WS, Heo DW, Maeng J, et al. Deep learning-based brain age prediction in patients with schizophrenia spectrum disorders. Schizophr Bull. 2024;50(4):804–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Sone D, Beheshti I, Tagai K, et al. Neuropsychiatric symptoms and neuroimaging based brain age in mild cognitive impairment and early dementia: a multicenter study. Psychiatry Clin Neurosci. 2025;79(4):158–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Iliff JJ, Lee H, Yu M, et al. Brain-wide pathway for waste clearance captured by contrast-enhanced MRI. J Clin Invest. 2013;123(3):1299–309. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Benveniste H, Liu X, Koundal S, Sanggaard S, Lee H, Wardlaw J. The glymphatic system and waste clearance with brain aging: a review. Gerontology. 2019;65(2):106–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Zhang X, He D. Glymphatic dysfunction across sleep disorders: a meta-analysis of DTI-ALPS studies. Front Neurol. 2025;17:1789842. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Xiong R, Feng J, Zhu H, et al. Evaluation of glymphatic system dysfunction in patients with insomnia via diffusion tensor image analysis along the perivascular space. Quant Imaging Med Surg. 2025;15(2):1114–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Cavaillès C, Dintica C, Habes M, Leng Y, Carnethon MR, Yaffe K. Association of self-reported sleep characteristics with neuroimaging markers of brain aging years later in middle-aged adults. Neurology. 2024;103(10):e209988. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Miao Y, Wang J, Li X, et al. Poor sleep health is associated with older brain age: the role of systemic inflammation. EBioMedicine. 2025;120:105941. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Chu C, Holst SC, Elmenhorst EM, et al. Total sleep deprivation increases brain age prediction reversibly in multisite samples of young healthy adults. J Neurosci. 2023;43(12):2168–77. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Clocchiatti-Tuozzo S, Rivier CA, Renedo D, et al. Life’s Essential 8 and poor brain health outcomes in middle-aged adults. Neurology. 2024;103(10):e209990. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Gosch Berton G, Lima Fonseca Rodrigues AC, Mombelli Mattei N, et al. The impact of chronic and acute sleep deprivation on key Alzheimer’s disease biomarkers: a systematic review and meta-analysis. Fluids Barriers CNS. 2026;23(1):31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57.Georgiopoulos C, Werlin A, Lasic S, et al. Diffusion tensor imaging along the perivascular space: the bias from crossing fibres. Brain Commun. 2024;6(6):fcae421. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58.Piantino JA, Iliff JJ, Lim MM, Levendovszky SR. Reader response: association of sleep, neuropsychological performance, and gray matter volume with glymphatic function in community-dwelling older adults. Neurology. 2023;100(7):355. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The raw data can be requested and downloaded via the official website of Cam-CAN project (https://www.cam-can.org/). The analyzed neuroimaging data that supports the findings of this study are available on request from the corresponding author.