Abstract
Brain glymphatic system function as measured by diffusion tensor image analysis along the perivascular space (DTI-ALPS) index has been associated with brain disorders, but its association with mortality and other diseases remains underexplored. This prospective study included 34,268 participants from the UK Biobank with MRI data collected since August 2014. The DTI-ALPS indices of the left and right hemispheres and the average index were calculated. The incident mortality and morbidity were defined as the death and disease (cancer, cardiovascular, respiratory, digestive, and neurodegenerative disease) reported between the MRI collection date and December 2021. Cox proportional hazard regression analyses were used to assess the association of DTI-ALPS indices with all-cause and cause-specific mortality and morbidity. We found that a higher DTI-ALPS index, indicated better glymphatic function, was associated with reduced risks of all-cause mortality (HR, 95% CI: 0.85, 0.78–0.93; P < 0.001), cardiovascular disease (CVD)-specific mortality (HR, 95% CI: 0.77, 0.63–0.93; P = 0.01), and neurodegenerative disease-specific mortality (HR, 95% CI: 0.56, 0.37–0.84; P = 0.01). Furthermore, an elevated DTI-ALPS index was associated with reduced risks of CVD (HR, 95% CI: 0.92, 0.89–0.96; P < 0.001) and neurodegenerative disease (HR, 95% CI: 0.68, 0.58–0.80; P < 0.001). The findings deepen the understanding of the role of brain glymphatic system in health.
Subject terms: Predictive markers, Diseases
Introduction
The glymphatic system is a brain-wide perivascular pathway parallel to the blood vasculature which facilitates the removal of extracellular solutes and cerebral waste products from the central nervous system through cerebrospinal fluid-interstitial fluid exchange [1, 2]. This system plays a key role in waste clearance including beta-amyloid and tau [1, 3], brain fluid regulation, and brain immunity [4]. Mounting evidence suggests that glymphatic failure is an important pathological pathway to a broad range of neurodegenerative diseases, including Alzheimer’s disease (AD) and Parkinson’s disease (PD) represented by pathologic protein aggregation [5]. Although the glymphatic system primarily functions to clear waste, it also serves as a mechanism for brain-wide signaling that could significantly impact an organism’s survival when adapting to a new environment [6]. It is a cornerstone in signaling among the brain, blood, gut, and immune system [6], and may thus act as an important pathological pathway to a broad range of diseases.
The diffusion tensor image analysis along the perivascular space (DTI-ALPS) index measured on MRI offers a non-invasive method to evaluate perivascular clearance activity in the human brain [7]. Prior studies have demonstrated a positive correlation between the DTI-ALPS index and glymphatic function, suggesting its value as a proxy for evaluating glymphatic performance in the human brain [8–10]. Furthermore, recent studies indicate that the DTI-ALPS index is linked to a variety of brain-related health outcomes, including cognition [11], sleep [12, 13], and neurodegenerative diseases [11, 14–16]. This index has also been associated with aging [17], white matter hyperintensities [18, 19], cardiovascular disease (CVD) [20–23], and renal function [24, 25]. Notably, peripheral metabolic factors have also been found to correlate with the DTI-ALPS index [8, 24]. Together, these results suggest the potential for the DTI-ALPS index to be associated with a broad range of health outcomes.
However, the findings from these previous studies primarily relied on cross-sectional designs, with only a few longitudinal cohort studies [26–30]. Longitudinal studies have mainly focused on specific diseases or outcomes, with no research conducted on the relationship between the DTI-ALPS index and health outcomes beyond those related to vascular and brain diseases. Particularly, the relationship between the DTI-ALPS index and mortality has not previously been well established in the literature. The aim of this prospective study was thus to assess whether glymphatic function as measured by the DTI-ALPS index is associated with all-cause and disease-specific mortality and morbidity. We found that a higher DTI-ALPS index, indicating better glymphatic function, was associated with a lower risk of all-cause mortality, and lower risks of cause-specific morbidity and mortality attributed to CVD and neurodegenerative disease.
Subjects and methods
Study sample
This prospective study used patient data from the UK Biobank. The UK Biobank is a population-based longitudinal cohort started in 2006 that includes data collected from 500,000 individuals aged 40–73 years. All participants were UK National Health Service registered and underwent comprehensive assessments at 22 centers across England, Scotland, and Wales (www.ukbiobank.ac.uk). Researchers using data from this database do not require separate ethical clearance and can operate under the Research Tissue Bank approval granted by the North West Multi-centre Research Ethics Committee.
The UK Biobank includes whole-body MRI data collected between August 2014 and October 2020. The current study included participants with MRI-DTI data and mortality information. Participants with preexisting conditions at the time of the initial MRI assessment were excluded from the morbidity analysis. Follow-up time was determined based either on the date of death (or date of first diagnosis) or when mortality data (or morbidity data) became available, whichever occurred first after MRI. Mortality data including the date of death and morbidity data including the date of initial diagnosis were updated until December 4th 2021 and November 1st 2021, respectively.
Morbidity and mortality outcomes
Morbidity and mortality outcomes were assessed using data acquired from NHS Digital in England and Wales or the NHS Central Register and National Records of Scotland. Disease diagnoses were classified according to the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10). The mortality was defined according to the underlying (primary) cause of death data. Morbidity and mortality codes for conditions including cancer, CVD, respiratory disease, digestive disease, and neurodegenerative disease are outlined in eMethods and eTable 1.
DTI-ALPS analysis
UK Biobank DTI imaging protocols, processing, and quality control are detailed in eMethods. The semi-automated DTI-ALPS pipeline was used (https://github.com/h8264/ALPS-index.git) [7]. DTIFIT (version 2.0; the FSL diffusion tensor fitting program, FMRIB) created diffusivity and fractional anisotropy (FA) maps for b = 1000. Two neurologists (Y.R.Z., 6 years; J.D., 7 years), blinded to clinical data, placed 4 mm spherical regions of interest (ROIs) on projection and association fibers in FA maps of 150 age- and gender-matched participants (see eFigure 1 for two examples). These ROI maps formed a 4D map, averaged into a 3D probability 60% probability map covering projection and association fibers. ROI maps were warped to Montreal Neurological Institute (MNI) space using FNIRT (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FNIRT) and binarized as templates. Individual diffusivity and FA maps were warped to MNI space. Diffusivity along the x, y, and z axes within the ROI was measured for projected and associated fibers. DTI-ALPS index was calculated for each hemisphere and averaged for a mean value.
Covariates
Patient age was determined from birthdates, while questionnaires identified sex, race, and Townsend Deprivation Index (TDI), a measure of socioeconomic status based on postal codes considering unemployment, car ownership, and congestion. Ethnicity information was collected and reported along with other demographic information without any implications. Other covariates included education level (categorized as college/university or other qualifications), body mass index (BMI), smoking history, alcohol consumption, and fruit-vegetable intake, which were assessed via touchscreen surveys. Physical activity was assessed via a modified version of the International Physical Activity Questionnaire. Physical activity, defined as moderate-to-vigorous, requires at least 150 minutes of moderate or 75 minutes of vigorous activity weekly. Fruit-vegetable intake was classified into <5, 5–8, and>8 servings/day.
Statistical analysis
Pearson correlation assessed the correlation between left and right DTI-ALPS indices. Data-driven quartile categorization was used: the mean, left, and right DTI-ALPS indices were divided into quartiles, with participant characteristics presented for each quartile (continuous variables: mean ± SD; categorical variables: percentages). Both continuous (per-SD) and restricted cubic spline models were performed in parallel to characterize the dose-response relationship without imposing arbitrary cut-offs.
Multivariable Cox proportional hazard model adjusted for age, sex, race, BMI, education, TDI, smoking history, alcohol intake history, and sleep disorders were incorporated to examine the association of DTI-ALPS index with mortality or morbidity risk. Restricted cubic spline analyses in multivariable Cox models assessed the relationship between all-cause mortality and the DTI-ALPS index, examining both linear (continuous values) and non-linear (categorical values divided by knots) associations. The time scale in the Cox model was defined as the period from the date of MRI collection to the date of death or disease diagnosis for incident events, and to the data update date of December 4th, 2021 for mortality controls or November 1st, 2021 for morbidity controls. In the main analyses of cause-specific mortality or morbidity, the competing causes were treated as censoring events. For cause-specific mortality analyses (e.g., CVD, neurodegenerative disease), cause-specific hazard models were used, in which deaths from competing causes were treated as censoring events at the time of death. The proportional hazards assumption was verified using the Schoenfeld residuals test for all fully adjusted models. The Cox results were represented as hazard ratio (HR) and 95% confidence interval (CI). Kaplan-Meier survival curves depicting all-cause mortality probability for each DTI-ALPS index quartile were compared using the Wald Chi-square test. The false discovery rate (FDR) method was corrected for multiple comparisons involved in the main analyses of the association of DTI-ALPS index as continuous variable with cause-specific mortality (3 DTI-ALPS indices * 5 outcomes) and morbidity (3 DTI-ALPS indices * 5 outcomes) respectively. Stratified analyses using multivariable Cox models were performed across categories of age, sex, race, BMI, education, smoking history, alcohol intake history, and sleep disorders status.
Sensitivity analyses were performed with multivariable Cox models by excluding early mortality and morbidity <2 years since MRI collection to reduce reverse causality, adjusting for additional factors, considering DTI-ALPS index quintiles, and excluding pre-diagnosed cases. Univariate and age-sex-adjusted Cox models were performed to reduce the possibility of model overfitting. Model discrimination was assessed using Harrell’s C-index, and internal validation was performed using bootstrap resampling (1,000 iterations) to estimate optimism-corrected C-index values. The Fine-Gray subdistribution hazard model accounted for competing mortality risks.
The threshold for statistical significance was set at nominal or FDR adjusted P < 0.05. R (version 4.3.1, The R Foundation) was used for all analyses and figure creation. Additional details and sample size justification are provided in eMethods.
Results
Patient characteristics
Of the 37,314 participants included in the UK Biobank whole-body MRI dataset, 3,046 were excluded for missing covariates (Fig. 1). The final study sample included 34,268 participants from the UK Biobank (mean age, 54.5 years ± 7.4 [SD]; 18,213 females, 16,055 males) among which 31,234 were White (91.1%), while 3034 were non-White (8.9%). Participants in Quartile 4 of the DTI-ALPS index appeared to be younger, less likely to smoke or drink alcohol, have a lower BMI, and experience fewer incident diseases (Table 1 and eTable 2). In addition, the left DTI-ALPS index was positively correlated with the right DTI-ALPS index (Pearson’s r = 0.82; P < 0.001) (eFig. 2).
Fig. 1. Flowchart and Study Design.
The left panel displays the flowchart of the study, which ultimately included 34,268 participants for the analysis of the association of DTI-ALPS with mortality. After excluding individuals with pre-diagnosed disease before MRI, the study incorporated 31,301 participants for cancer, 25,738 participants for CVD, 30,457 participants for respiratory disease, 22,796 participants for digestive disease, and 34,222 participants for neurodegeneration disease, for the analyses of the associations of DTI-ALPS with morbidity. The top right panel illustrates the computation of DTI-ALPS within this study. The bottom right panel demonstrates the use of multivariable Cox proportional hazards models to investigate the relationships between DTI-ALPS, mortality, and morbidity, adjusting for age, sex, race, BMI, educational levels, TDI, smoking history, alcohol intake history, and sleep disorders. BMI = body mass index, DTI-ALPS = diffusion tensor image analysis along the perivascular space, TDI = Townsend Deprivation Index.
Table 1.
Baseline characteristics of study participants stratified by mean DTI-ALPS index quartiles.
| Overall | Mean DTI-ALPS index | ||||
|---|---|---|---|---|---|
| Quartile 1 (0.77–1.27) | Quartile 2 (1.27–1.35) | Quartile 3 (1.25–1.44) | Quartile 4 (1.44–1.88) | ||
| N | 34268 | 8567 | 8567 | 8567 | 8567 |
| Age (y) | 54.5 ± 7.4 | 57.0 ± 7.2 | 54.9 ± 7.4 | 53.7 ± 7.2 | 52.3 ± 6.9 |
| Sex | |||||
| Female | 18213 (53.1) | 3501 (40.9) | 4282 (50.0) | 4888 (57.1) | 5542 (64.7) |
| Male | 16055 (46.9) | 5066 (59.1) | 4285 (50.0) | 3679 (42.9) | 3025 (35.3) |
| Education | |||||
| College/University | 17076 (49.8) | 4345 (50.7) | 4259 (49.7) | 4251 (49.6) | 4221 (49.3) |
| Other qualifications | 17192 (50.2) | 4222 (49.3) | 4308 (50.3) | 4316 (50.4) | 4346 (50.7) |
| Race | |||||
| White | 31234 (91.1) | 7919 (92.4) | 7833 (91.4) | 7773 (90.7) | 7709 (90.0) |
| Non-white | 3034 (8.9) | 648 (7.6) | 734 (8.6) | 794 (9.3) | 858 (10.0) |
| Townsend deprivation Index | −1.9 ± 2.7 | −2.0 ± 2.7 | −1.9 ± 2.7 | −2.0 ± 2.7 | −1.9 ± 2.7 |
| BMI (kg/m2) | 26.5 ± 4.2 | 26.9 ± 4.2 | 26.5 ± 4.1 | 26.4 ± 4.2 | 26.1 ± 4.2 |
| MVPA | |||||
| No | 13803 (46.2) | 3487 (46.5) | 3385 (45.4) | 3443 (46.3) | 3488 (46.8) |
| Yes | 16048 (53.8) | 4017 (53.5) | 4072 (54.6) | 3996 (53.7) | 3963 (53.2) |
| Fruit & vegetable Intake (servings/day) | |||||
| <5 | 27556 (87.1) | 6865 (86.7) | 6854 (87.0) | 6886 (86.8) | 6951 (87.7) |
| 5–8 | 2534 (8.0) | 652 (8.2) | 621 (7.9) | 642 (8.1) | 619 (7.8) |
| >8 | 1564 (4.9) | 404 (5.1) | 400 (5.1) | 401 (5.1) | 359 (4.5) |
| Smoking history | |||||
| Never | 21246 (62.0) | 4981 (58.1) | 5243 (61.2) | 5369 (62.7) | 5653 (66.0) |
| Former | 10961 (32.0) | 3018 (35.2) | 2798 (32.7) | 2686 (31.4) | 2459 (28.7) |
| Current | 2061 (6.0) | 568 (6.6) | 526 (6.1) | 512 (6.0) | 455 (5.3) |
| Alcohol intake history | |||||
| Never drink | 806 (2.4) | 178 (2.1) | 193 (2.3) | 206 (2.4) | 229 (2.7) |
| Previous drink | 697 (2.0) | 197 (2.3) | 144 (1.7) | 178 (2.1) | 178 (2.1) |
| Current drink | 32765 (95.6) | 8192 (95.6) | 8230 (96.1) | 8183 (95.5) | 8160 (95.2) |
Continuous variables are presented as mean ± SD. Categorical variables are presented as the number of participants, with percentages in parentheses. BMI values were defined based on height and weight measured during the initial Assessment Centre visit. Townsend deprivation index was defined as a score assigned to each participant based on postal codes considering unemployment, car ownership, and congestion. MVPA (yes/no) was defined as whether at least 150 min of moderate-intensity activity weekly, 75 min of vigorous activity weekly, or a proportional combination.
BMI, body mass index, DTI-ALPS, diffusion tensor image analysis along the perivascular space, MVPA, moderate-to-vigorous physical activity.
Higher DTI-ALPS index was associated with reduced risk of mortality
A total of 551 of 34,268 (1.6%) participants died in a median 4.58-year follow-up (interquartile range [IQR], 3.78–6.06 years; 170,545 person-years. 316 participants died from cancer, 117 from CVD, 11 from digestive diseases, 20 from respiratory diseases, and 27 from neurodegenerative diseases. The multivariable Cox proportional hazard models showed that higher DTI-ALPS index was associated with reduced risks of all-cause mortality (mean DTI-ALPS index, [HR, 95% CI: 0.85, 0.78–0.93; P < 0.001]; left DTI-ALPS index, [HR, 95% CI: 0.86, 0.78–0.94; P < 0.001]; right DTI-ALPS index, [HR, 95% CI: 0.86, 0.78–0.93; P < 0.001]). The lowest quartile for mean DTI-ALPS index (HR, 95% CI: 1.51, 1.16–1.98; P < 0.003) and left DTI-ALPS index (HR, 95% CI: 1.35, 1.03- 1.76; P = 0.03) were found to be associated with a higher risk of all-cause mortality compared to the highest quartile (Table 2). For the right DTI-ALPS index, all quartiles had a higher risk of all-cause mortality compared to the highest quartile (HR range, 1.33–1.47; P range, 0.005–0.048). Furthermore, a higher mean DTI-ALPS index showed a negative correlation with risks for all-cause mortality (P = 0.002). Similarly, the left DTI-ALPS index (P = 0.002) and right DTI-ALPS index (P = 0.004) were negatively correlated with risks for all-cause mortality (Fig. 2). Comparisons of Kaplan-Meier curves indicate that the lowest quartile of the DTI-ALPS index is associated with the highest probability of mortality. Furthermore, there is a gradual increase of mortality risk as the DTI-ALPS index quartile decreases (all Wald χ2 P < 0.001) (Fig. 2).
Table 2.
Cox Proportional Hazard analyses results for the association of DTI-ALPS index with all-cause mortality.
| DTI-ALPS index | Number of participants | Cases/Control | HR (95% CI) | P value | P value for quartile trend |
|---|---|---|---|---|---|
| Mean DTI-ALPS index (continuous, 0.77–1.88) | 34268 | 551/33717 | 0.85 (0.78, 0.93) | <0.001 | |
| Quartile 4 (1.44–1.88) | 8567 | 79/8488 | 1 [Reference] | – | 0.004 |
| Quartile 3 (1.35–1.44) | 8567 | 124/8443 | 1.28 (0.97, 1.70) | 0.09 | |
| Quartile 2 (1.27–1.35) | 8567 | 142/8425 | 1.28 (0.97, 1.70) | 0.08 | |
| Quartile 1 (0.77–1.27) | 8567 | 206/8361 | 1.51 (1.16, 1.98) | 0.003 | |
| Left DTI-ALPS index (continuous, 0.77–1.90) | 34268 | 551/33717 | 0.86 (0.78, 0.94) | <0.001 | |
| Quartile 4 (1.46–1.90) | 8567 | 84/8483 | 1 [Reference] | – | 0.03 |
| Quartile 3 (1.37–1.46) | 8567 | 121/8446 | 1.19 (0.90, 1.58) | 0.22 | |
| Quartile 2 (1.28–1.37) | 8567 | 148/8419 | 1.25 (0.95, 1.64) | 0.11 | |
| Quartile 1 (0.77–1.28) | 8567 | 198/8369 | 1.35 (1.03, 1.76) | 0.03 | |
| Right DTI-ALPS index (continuous, 0.72–1.98) | 34268 | 551/33717 | 0.86 (0.78, 0.93) | <0.001 | |
| Quartile 4 (1.43–1.98) | 8567 | 80/8487 | 1 [Reference] | – | 0.009 |
| Quartile 3 (1.33–1.43) | 8567 | 126/8441 | 1.33 (1.00, 1.76) | 0.048 | |
| Quartile 2 (1.24–1.33) | 8565 | 152/8413 | 1.40 (1.06, 1.84) | 0.02 | |
| Quartile 1 (0.72–1.24) | 8569 | 193/8376 | 1.47 (1.12, 1.92) | 0.005 |
Data in parentheses are 95% CIs. Hazard ratios and P value were estimated from multivariable Cox proportional hazard models. The P value for the quartile trend was estimated from the multivariable Cox proportional hazard model, treating the DTI-ALPS indices as an ordinal variable (quantiles 1 to 4 are labeled 1 to 4). All models were adjusted for age, sex, race, BMI, education, TDI, smoking history, alcohol intake history, and sleep disorders.
BMI, body mass index, DTI-ALPS, diffusion tensor image analysis along the perivascular space, HR, hazard ratio, CI, confidence interval, TDI, Townsend Deprivation Index.
Fig. 2. The association of DTI-ALPS index with all-cause mortality.
The associations of all-cause mortality with the mean (A), left (B), and right (C) DTI-ALPS indices were evaluated. The plots on the left show the negative correlation between the DTI-ALPS index and all-cause mortality. This plot was created using restricted cubic splines in Cox proportional hazard models with knots at the 10th, 50th, and 90th percentiles of the exposure distribution, and the P values for linear association were reported. The shaded areas denote 95% CIs. Plots in the middle show Cox proportional hazard ratios for all-cause mortality stratified by DTI-ALPS index quartiles. Points represent the HR values, and whiskers represent the 95% CIs. Plots on the right show Kaplan-Meier survival curves illustrating the population-averaged all-cause survival probabilities for each quartile of the DTI-ALPS index. Survival curves were compared using a Wald χ2 test. All models were adjusted for age, sex, race, BMI, educational levels, TDI, smoking history, alcohol intake history, and sleep disorders. BMI = body mass index, DTI-ALPS = diffusion tensor image analysis along the perivascular space, TDI = Townsend Deprivation Index.
Higher DTI-ALPS indices were associated with decreased risks of CVD-specific mortality (HR range, 0.77–0.79; all P values = 0.01) and neurodegenerative disease-specific mortality (HR range, 0.55–0.60; P range, 0.003-0.02) (Fig. 3). The results remained significant after correction for multiple comparisons (eTable 3). No evidence of the association of DTI-ALPS index with the risk of cancer, digestive disease, and respiratory disease-related mortality was observed (Fig. 3 and eTable 3). The multivariable Cox proportional hazard models showed an association of increased quartiles of DTI-ALPS index with decreased risk of CVD-specific mortality (P range for quartile trend, 0.033–0.048) and neurodegenerative disease-specific mortality (P range for quartile trend, <0.001–0.01) (eTable 4, 5).
Fig. 3. The association of DTI-ALPS index with cause-specific mortality.
The forest plot shows the results of the Cox proportional hazards regression models used to evaluate the association of the mean, left, and right DTI-ALPS indices with the risk of cause-specific mortality. All models were adjusted for age, sex, race, BMI, educational levels, TDI, smoking history, alcohol intake history, and sleep disorders. BMI = body mass index, DTI-ALPS = diffusion tensor image analysis along the perivascular space, TDI = Townsend Deprivation Index.
Higher DTI-ALPS index was associated with reduced risk of morbidity
After excluding individuals diagnosed before the initial MRI assessment, there were 2,023 diagnosed with cancer (median follow-up 4.41 years [IQR, 2.61–5.87 years; 149,106 person-years]), 3,390 with CVD (4.29 years [IQR, 3.50–5.63 years; 117,966 person-years]), 3,650 with digestive disease (4.24 years [IQR, 3.47–5.57 years; 102,669 person-years]), 1,944 with respiratory disease (4.41 years [IQR, 3.61–5.82 years; 144,927 person-years]), and 166 with neurodegenerative disease (4.5 years [IQR, 3.70–6.00 years; 168,130 person-years]). An elevated DTI-ALPS index was associated with reduced risk of CVD (mean DTI-ALPS index, [HR, 95% CI: 0.92, 0.89–0.96; P < 0.001]; left DTI-ALPS index, [HR, 95% CI: 0.93, 0.89–0.96; P < 0.001]; right DTI-ALPS index, [HR, 95% CI: 0.93, 0.90–0.96; P < 0.001]) and neurodegenerative (mean DTI-ALPS index, [HR, 95% CI: 0.68, 0.58–0.80; P < 0.001]; left DTI-ALPS index, [HR, 95% CI: 0.66, 0.56–0.78; P < 0.001]; right DTI-ALPS index, [HR, 95% CI: 0.72, 0.61–0.85; P < 0.001]) morbidity (Fig. 4). The results remained significant after correction for multiple comparisons (eTable 6). No evidence of an association of the DTI-ALPS index with the risk of cancer, digestive, or respiratory disease morbidity was observed. Furthermore, the increased DTI-ALPS index quartiles were associated with reduced risk of CVD (all P for quartile trend < 0.001), neurodegenerative (all P for quartile trend < 0.001), and digestive (P range for quartile trend, 0.002–0.001) morbidity (eTable 7, 8).
Fig. 4. The association of DTI-ALPS index with cause-specific morbidity.
The forest plot shows the results of the Cox proportional hazards regression models used to evaluate the association of the mean, left, and right DTI-ALPS indices with the risk of cause-specific morbidity. All models were adjusted for age, sex, race, BMI, educational levels, TDI, smoking history, alcohol intake history, and sleep disorders. BMI = body mass index, DTI-ALPS = diffusion tensor image analysis along the perivascular space, TDI = Townsend Deprivation Index.
Subgroup analysis
Both patients ≤55 years (HR range, 0.80–0.81; P range, 0.017–0.023) and >55 years (HR range, 0.82–0.83; all P values < 0.001) of age showed an association of DTI-ALPS index with reduced risk of all-cause mortality. However, for other subgroups, multivariable Cox proportional hazard models showed this association only existed in specific subgroups, including subgroups of male (HR range, 0.81–0.82; all P values < 0.001), white race (HR range, 0.85–0.86; all P values < 0.001), former (HR range, 0.83–0.84; P range, 0.01–0.017) and current smoking (HR range, 0.64–0.70; P range, 0.002–0.017), current alcohol intake (HR range, 0.84–0.85; all P values < 0.001), education level of other qualifications (HR range, 0.81–0.84; P range, <0.001–0.005), BMI within 25–29.9 (HR range, 0.85–0.87; P range, 0.017–0.038), BMI ≥ 30 (HR range, 0.78–0.81; P range, 0.008–0.022), and no sleep disorders (HR range, 0.85–0.86; all P values < 0.001) (eTable 9).
Sensitivity analysis
The decreased risk of mortality and morbidity associated with increased DTI-ALPS index persisted when excluding early deaths/morbidity <2 years (eTable 10, 11) and after adjusting for chronic disease and lifestyle factors (eTable 12, 13). With no covariate adjustment or adjusting only for age/sex, the negative association of DTI-ALPS with CVD and neurodegenerative disease mortality remained, and the negative cancer association became significant (eTable 14, 15). Fine-Gray models showed elevated DTI-ALPS associated with reduced CVD (HR range, 0.77–0.79; P range, 0.008–0.012) and neurodegenerative disease (HR range, 0.55–0.6; P range, 0.003–0.001) mortality after accounting for the competing risks (eTable 16). Excluding pre-diagnosed cases, the DTI-ALPS index was associated with decreased neurodegenerative (HR range, 0.55–0.59; P range, 0.004–0.012) but not CVD mortality risk (eTable 17). Comparing quintiles, the lowest DTI-ALPS quintile was associated with higher all-cause mortality vs. the highest quintile (HR range, 1.52–1.64; P range, 0.004–0.014) (eTable 18).The fully adjusted models demonstrated good discrimination for all-cause mortality (C-index = 0.730, optimism-corrected C-index = 0.725) and neurodegenerative disease incidence (C-index = 0.764, optimism-corrected C-index = 0.749), with negligible optimism (range, 0.002–0.033), indicating minimal overfitting (eTable 19).
Discussion
Glymphatic function as measured by DTI-ALPS index on MRI has been associated with AD and PD, but its association with other causes of mortality and morbidity remains underexplored. Thus, we aimed to assess whether glymphatic function as measured by the DTI-ALPS index is associated with all-cause and disease-specific mortality and morbidity. We found that a higher DTI-ALPS index was associated with lower risks of all-cause mortality as well as mortality caused by CVD and neurodegenerative disease. Additionally, an elevated DTI-ALPS index was associated with reduced risks of morbidity caused by CVD and neurodegenerative disease.
The prior cross-sectional study linked a high DTI-ALPS index to increased vascular pathology and reduced CVD risk [21]. Similarly, we found a higher DTI-ALPS index associated with reduced risks of CVD-related mortality and morbidity. The association between the DTI-ALPS index and CVD risk may be explained by the close coupling between vascular dynamics and glymphatic function. Cerebral arterial pulsatility serves as a primary driving force for CSF influx into the brain parenchyma along perivascular spaces [31]. Vascular risk factors such as hypertension and reduced arterial compliance may impair this pulsatility-driven CSF-ISF circulation, leading to glymphatic dysfunction [32, 33]. Prior work has positioned the glymphatic system as a critical mediator in the heart-brain axis disease continuum [34]. Emerging evidence indicates that the glymphatic system is functionally connected to the meningeal lymphatic system, and together these pathways facilitate the drainage of brain-derived solutes to the cervical lymph nodes [2]. Solutes originating from the brain can be transported via this route to peripheral lymphoid tissues, where they may activate systemic inflammatory responses, accelerating endothelial dysfunction and the formation and rupture of atherosclerotic plaques. Moreover, CVD-related glymphatic dysfunction mediates its deleterious effects on brain health (e.g., cognitive impairment) [31, 35] and may create a vicious cycle of glymphatic-cardiovascular injury, thereby increasing CVD mortality risk.
Additionally, we also found that a higher DTI-ALPS index was associated with reduced neurodegenerative disease-related mortality and morbidity risks. Previous studies showed that a decreased DTI-ALPS index correlates with cognitive decline and increased AD pathology [14]. Various neurodegenerative diseases, including AD, PD [15, 27, 28], and amyotrophic lateral sclerosis [16], have also been found to exhibit lower DTI-ALPS indices than healthy controls. The association between the DTI-ALPS index and neurodegenerative disease risk may be mediated by the glymphatic system’s capacity to clear neurotoxic proteins. A reduced DTI-ALPS index reflects diminished perivascular clearance capacity, potentially leading to decreased elimination of neurotoxic proteins such as Aβ, tau, and α-synuclein from the brain parenchyma [36]. The consequent accumulation and aggregation of these proteins may promote their progressive deposition and cell-to-cell propagation, directly exacerbating neuronal injury and synaptic dysfunction, thereby accelerating disease progression [36]. Furthermore, the aggregation of misfolded proteins may induce astrocytic and microglial activation, releasing inflammatory mediators that could further impair perivascular clearance mechanisms. Thus, the glymphatic system may represent a common pathway in the pathogenesis of neurodegenerative diseases [5], with the DTI-ALPS index serving as a sensitive indicator of this waste clearance capacity.
We also observed a negative association of the quartile of the DTI-ALPS index with digestive disease diagnosis risk. However, no association with digestive disease mortality risk was found, potentially due to the limited number of digestive disease mortality (N = 11) in this study. Previous research links impaired brain glymphatic clearance to diseases like hepatic encephalopathy exhibiting harmful brain metabolite accumulation [37, 38]. Additionally, the potential exists for glymphatic system cross-talk with the gastrointestinal tract via vagus nerve retrograde signaling or blood-brain barrier molecule crossing [4]. Furthermore, in unadjusted and age-sex-adjusted models, we found a negative association of DTI-ALPS with cancer mortality. Prior studies linked DTI-ALPS to glioma [39, 40], but the mechanism is unclear. Further research is needed to validate these associations and elucidate underlying mechanisms.
Our study revealed that the DTI-ALPS index is associated with reduced risk of all-cause mortality in specific subgroups such as males, individuals who currently or formerly smoked, and individuals who currently drink alcohol, though the underlying mechanism remains uncertain. Males have been found to have a lower DTI-ALPS index than females [22], implying that sex-related biological pathways may be involved in the glymphatic system. Alternatively, glymphatic function may counteract the adverse effects of these factors. For example, the DTI-ALPS index is negatively correlated with obesity and smoking history [24].
This study has several limitations. First, the lack of standardized ROI segmentation/sizing criteria may raise potential selection bias from our 4 mm ROI radius choice for DTI-ALPS index calculation. Future research should establish uniform guidelines for standardized DTI-ALPS calculation. Second, as an observational study, we cannot establish a causal link between the DTI-ALPS index and mortality or morbidity. Third, cancer accounted for 57% of all-cause mortality may introduce bias. Fourth, over 90% of participants in the present study may limit the generalizability of our findings to other racial and ethnic groups. Fifth, relatively low neurodegenerative disease incidence in mortality/morbidity analyses may limit the clinical utility of the finding. Sixth, unmeasured factors like inflammation may influence the DTI-ALPS index-outcomes association. Seventh, our study found inconsistent results for the DTI-ALPS index in the left and right hemispheres. While the source of this inconsistency is unclear, we cannot entirely exclude the possibility that these disparities might be due to computational errors. Finally, while DTI-ALPS shows promise as a glymphatic function marker, it requires further validation.
In conclusion, an elevated DTI-ALPS index was associated with reduced all-cause mortality as well as CVD and neurodegenerative disease-specific mortality and morbidity, suggesting potential as a clinical tool for improved health management. If validated as an accurate glymphatic function surrogate, the DTI-ALPS index could have prospective clinical implications for the treatment of CVD and neurodegenerative disease. However, this nascent area necessitates comprehensive further investigation before clinical implementation.
Supplementary information
Acknowledgements
The authors thank the participants, contributors, and researchers of UK Biobank for making the data available. This work is supported by grants from the National Key Research and Development Program of China (2023YFC3605400), China National Postdoctoral Program for Innovative Talents (BX20230089), China Postdoctoral Science Foundation (2023M740686), Shanghai Municipal Health Commission Health Industry Clinical Research Special Project (20244Y0124), and ZHANGJIANG LAB, Tianqiao and Chrissy Chen Institute, and the State Key Laboratory of Neurobiology and Frontiers Center for Brain Science of Ministry of Education, Fudan University, China.
Author contributions
YZ, ZW, and SH contributed equally. YZ, ZW, SH, and JD had full access to all the data in the study and took responsibility for the integrity of the data and the accuracy of the data analysis. Concept and design: JY. Administrative, technical, or material support: JF and WC. Acquisition, statistical analysis, or interpretation of data: YZ, ZW, SH, JD, YF, PG. Biostatistical analysis technical support: WC. Drafting of the manuscript: YZ, ZW. Critical revision of the manuscript for important intellectual content: All authors. Supervision: JY.
Data availability
The data reported in this study cannot be deposited in a public repository because these data were accessed under a specific license agreement of UK Biobank, precluding their public availability. To request access, contact UK Biobank access management team, detailed at https://www.ukbiobank.ac.uk/register-apply/. For this study, data were procured via application number 19542 from the UK Biobank repository.
Code availability
Package survival in R version 4.3.1 were used to perform Cox proportional hazards regression model. Scripts used to perform the analyses are available at https://github.com/wwwhales/ukb-alps-mortality.
Competing interests
The authors declare no competing interests.
Ethics approval and consent to participate
All participants gave written informed consent prior data collection. The UK Biobank has ethical approval by the National Research Ethics Service (Ref 11/NW/0382).
Footnotes
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
These authors contributed equally: Ya-Ru Zhang, Zhi-Bo Wang, Shu-Yi Huang.
Supplementary information
The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04381-z.
References
- 1.Iliff JJ, Wang M, Liao Y, Plogg BA, Peng W, Gundersen GA, et al. A paravascular pathway facilitates CSF flow through the brain parenchyma and the clearance of interstitial solutes, including amyloid β. Sci Transl Med. 2012;4:147ra11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Louveau A, Smirnov I, Keyes TJ, Eccles JD, Rouhani SJ, Peske JD, et al. Structural and functional features of central nervous system lymphatic vessels. Nature. 2015;523:337–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Iliff JJ, Chen MJ, Plog BA, Zeppenfeld DM, Soltero M, Yang L, et al. Impairment of glymphatic pathway function promotes tau pathology after traumatic brain injury. J Neurosci. 2014;34:16180–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Hablitz LM, Nedergaard M. The glymphatic system: a novel component of fundamental neurobiology. J Neurosci. 2021;41:7698–711. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Nedergaard M, Goldman SA. Glymphatic failure as a final common pathway to dementia. Science. 2020;370:50–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Lohela TJ, Lilius TO, Nedergaard M. The glymphatic system: implications for drugs for central nervous system diseases. Nat Rev Drug Discov. 2022;21:763–79. [DOI] [PubMed] [Google Scholar]
- 7.Taoka T, Masutani Y, Kawai H, Nakane T, Matsuoka K, Yasuno F, 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;35:172–8. [DOI] [PubMed] [Google Scholar]
- 8.Andica C, Kamagata K, Takabayashi K, Kikuta J, Kaga H, Someya Y, et al. Neuroimaging findings related to glymphatic system alterations in older adults with metabolic syndrome. Neurobiol Dis. 2023;177:105990. [DOI] [PubMed] [Google Scholar]
- 9.Liu X, Barisano G, Shao X, Jann K, Ringman JM, Lu H, et al. Cross-vendor test-retest validation of diffusion tensor image analysis along the perivascular space (DTI-ALPS) for evaluating glymphatic system function. Aging Dis. 2024;15:1885–98. [DOI] [PMC free article] [PubMed]
- 10.Huang SY, Zhang YR, Guo Y, Du J, Ren P, Wu BS, et al. Glymphatic system dysfunction predicts amyloid deposition, neurodegeneration, and clinical progression in Alzheimer’s disease. Alzheimers Dement. 2024;20:3251–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Kamagata K, Andica C, Takabayashi K, Saito Y, Taoka T, Nozaki H, et al. Association of MRI indices of glymphatic system with amyloid deposition and cognition in mild cognitive impairment and Alzheimer disease. Neurology. 2022;99:e2648–e60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Siow TY, Toh CH, Hsu J-L, Liu G-H, Lee S-H, Chen N-H, et al. Association of sleep, neuropsychological performance, and gray matter volume with glymphatic function in community-dwelling older adults. Neurology. 2022;98:e829–e38. [DOI] [PubMed] [Google Scholar]
- 13.Gumeler E, Aygun E, Tezer FI, Saritas EU, Oguz KK. Assessment of glymphatic function in narcolepsy using DTI-ALPS index. Sleep Med. 2023;101:522–7. [DOI] [PubMed] [Google Scholar]
- 14.Hsu J-L, Wei Y-C, Toh CH, Hsiao I-T, Lin K-J, Yen T-C, et al. Magnetic resonance images implicate that glymphatic alterations mediate cognitive dysfunction in Alzheimer disease. Ann Neurol. 2023;93:164–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Shen T, Yue Y, Ba F, He T, Tang X, Hu X, et al. Diffusion along perivascular spaces as marker for impairment of glymphatic system in Parkinson’s disease. NPJ Parkinson’s Dis. 2022;8:174. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Liu S, Sun X, Ren Q, Chen Y, Dai T, Yang Y, et al. Glymphatic dysfunction in patients with early-stage amyotrophic lateral sclerosis. Brain. 2024;147:100–8. [DOI] [PubMed]
- 17.Dai Z, Yang Z, Chen X, Zheng W, Zhuang Z, Liao Y, et al. The aging of glymphatic system in human brain and its correlation with brain charts and neuropsychological functioning. Cereb Cortex. 2023;33:7896–903. [DOI] [PubMed] [Google Scholar]
- 18.Li Y, Zhou Y, Zhong W, Zhu X, Chen Y, Zhang K, et al. Choroid plexus enlargement exacerbates white matter hyperintensity growth through glymphatic impairment. Ann Neurol. 2023;94:182–95. [DOI] [PubMed] [Google Scholar]
- 19.Cai J, Sun J, Chen H, Chen Y, Zhou Y, Lou M, et al. Different mechanisms in periventricular and deep white matter hyperintensities in old subjects. Front Aging Neurosci. 2022;14:940538. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Toh CH, Siow TY. Glymphatic dysfunction in patients with ischemic stroke. Front Aging Neurosci. 2021;13:756249. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Tian Y, Cai X, Zhou Y, Jin A, Wang S, Yang Y, et al. Impaired glymphatic system as evidenced by low diffusivity along perivascular spaces is associated with cerebral small vessel disease: a population-based study. Stroke Vasc Neurol. 2023;8:413–23. [DOI] [PMC free article] [PubMed]
- 22.Zhang Y, Zhang R, Ye Y, Wang S, Jiaerken Y, Hong H, et al. The influence of demographics and vascular risk factors on glymphatic function measured by diffusion along perivascular space. Front Aging Neurosci. 2021;13:693787. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Kikuta J, Kamagata K, Takabayashi K, Taoka T, Yokota H, Andica C, et al. An investigation of water diffusivity changes along the perivascular space in elderly subjects with hypertension. AJNR Am J Neuroradiol. 2022;43:48–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Wei Y-C, Hsu C-CH, Huang W-Y, Lin C, Chen C-K, Chen Y-L, et al. Vascular risk factors and astrocytic marker for the glymphatic system activity. La Radiologia Medica. 2023. [DOI] [PMC free article] [PubMed]
- 25.Heo CM, Lee WH, Park BS, Lee YJ, Park S, Kim YW, et al. Glymphatic dysfunction in patients with end-stage renal disease. Front Neurol. 2022;12:809438. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Wang J, Zhou Y, Zhang K, Ran W, Zhu X, Zhong W, et al. Glymphatic function plays a protective role in ageing-related cognitive decline. Age Ageing. 2023;52:afad107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Miljkovic-Gacic I, Wang X, Kammerer CM, Bunker CH, Patrick AL, Wheeler VW, et al. Sex and genetic effects on upper and lower body fat and associations with diabetes in multigenerational families of African heritage. Metabolism. 2008;57:819–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Si X, Guo T, Wang Z, Fang Y, Gu L, Cao L, et al. Neuroimaging evidence of glymphatic system dysfunction in possible REM sleep behavior disorder and Parkinson’s disease. NPJ Parkinson’s Dis. 2022;8:54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Xu J, Su Y, Fu J, Wang X, Nguchu BA, Qiu B, et al. Glymphatic dysfunction correlates with severity of small vessel disease and cognitive impairment in cerebral amyloid angiopathy. Eur J Neurol. 2022;29:2895–904. [DOI] [PubMed] [Google Scholar]
- 30.Bae YJ, Kim J-M, Choi BS, Ryoo N, Song YS, Nam Y, et al. Altered brain glymphatic flow at diffusion-tensor MRI in rapid eye movement sleep behavior disorder. Radiology. 2023;307:e221848. [DOI] [PubMed] [Google Scholar]
- 31.Kritsilis M, Vanherle L, Rosenholm M, In ‘t Zandt R, Yao Y, Swanberg KM, et al. Loss of glymphatic homeostasis in heart failure. Brain. 2025;148:985–1000. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Iliff JJ, Wang M, Zeppenfeld DM, Venkataraman A, Plog BA, Liao Y, et al. Cerebral arterial pulsation drives paravascular CSF-interstitial fluid exchange in the murine brain. J Neurosci. 2013;33:18190–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Mestre H, Tithof J, Du T, Song W, Peng WG, Sweeney AM, et al. Flow of cerebrospinal fluid is driven by arterial pulsations and is reduced in hypertension. Nat Commun. 2018;9:4878. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Zheng C, Cui Y, Qin R, Si J, Xiao K, Li G, et al. Association of glymphatic system dysfunction with cardiac injury and cognitive impairment in heart failure: a multimodal MRI study. Acad Radiol. 2025;32:3227–36. [DOI] [PubMed] [Google Scholar]
- 35.Guo J, Zhang Z, Meng X, Jing J, Hu Y, Yao Y, et al. Atrial fibrillation catheter ablation, brain glymphatic function, and cognitive performance. Eur Heart J. 2025;46:1733–43. [DOI] [PubMed] [Google Scholar]
- 36.Lopes DM, Llewellyn SK, Harrison IF. Propagation of tau and α-synuclein in the brain: therapeutic potential of the glymphatic system. Transl Neurodegener. 2022;11:19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Hadjihambi A, Harrison IF, Costas-Rodríguez M, Vanhaecke F, Arias N, Gallego-Durán R, et al. Impaired brain glymphatic flow in experimental hepatic encephalopathy. J Hepatol. 2019;70:40–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Hsu S-J, Zhang C, Jeong J, Lee S-I, McConnell M, Utsumi T, et al. Enhanced meningeal lymphatic drainage ameliorates neuroinflammation and hepatic encephalopathy in cirrhotic rats. Gastroenterology. 2021;160:1315–29.e13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Zhu H, Xie Y, Li L, Liu Y, Li S, Shen N, et al. Diffusion along the perivascular space as a potential biomarker for glioma grading and isocitrate dehydrogenase 1 mutation status prediction. Quant Imaging Med Surg. 2023;13:8259–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Toh CH, Siow TY. Factors associated with dysfunction of glymphatic system in patients with glioma. Front Oncol. 2021;11:744318. [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 data reported in this study cannot be deposited in a public repository because these data were accessed under a specific license agreement of UK Biobank, precluding their public availability. To request access, contact UK Biobank access management team, detailed at https://www.ukbiobank.ac.uk/register-apply/. For this study, data were procured via application number 19542 from the UK Biobank repository.
Package survival in R version 4.3.1 were used to perform Cox proportional hazards regression model. Scripts used to perform the analyses are available at https://github.com/wwwhales/ukb-alps-mortality.