Perivascular spaces, plasma GFAP, and speeded executive function in neurodegenerative diseases

Abstract

INTRODUCTION

We investigated the effect of perivascular spaces (PVS) volume on speeded executive function (sEF), as mediated by white matter hyperintensities (WMH) volume and plasma glial fibrillary acidic protein (GFAP) in neurodegenerative diseases.

METHODS

A mediation analysis was performed to assess the relationship between neuroimaging markers and plasma biomarkers on sEF in 333 participants clinically diagnosed with Alzheimer's disease/mild cognitive impairment, frontotemporal dementia, or cerebrovascular disease from the Ontario Neurodegenerative Disease Research Initiative.

RESULTS

PVS was significantly associated with sEF (c = ‐0.125 ± 0.054, 95% bootstrap confidence interval [CI] [‐0.2309, ‐0.0189], p = 0.021). This effect was mediated by both GFAP and WMH.

DISCUSSION

In this unique clinical cohort of neurodegenerative diseases, we demonstrated that the effect of PVS on sEF was mediated by the presence of elevated plasma GFAP and white matter disease. These findings highlight the potential utility of imaging and plasma biomarkers in the current landscape of therapeutics targeting dementia.

Highlights

• Perivascular spaces (PVS) and white matter hyperintensities (WMH) are imaging markers of small vessel disease.

• Plasma glial fibrillary protein acidic protein (GFAP) is a biomarker of astroglial injury.

• PVS, WMH, and GFAP are relevant in executive dysfunction from neurodegeneration.

• PVS's effect on executive function was mediated by GFAP and white matter disease.

1. INTRODUCTION

Under normal conditions, the perivascular spaces (PVS) surrounding the brain's arterioles, capillaries, and venules, are an important component of the glymphatic system. ¹ , ² This perivascular system of tunnels serves as a fluid pathway through which interstitial fluids circulate and metabolic waste solutes can be cleared out of the brain. ³ , ⁴ Normal PVS are not typically seen on standard structural MRI and can only be visualized when enlarged. Thus, MRI‐visible PVS, often observed in cerebrovascular and neurodegenerative disease patients, ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² are believed to be an indicator of glymphatic dysfunction and are recently recognized as an imaging‐based feature of cerebral small vessel disease. ¹³

It is unclear, however, if PVS burden is associated with cognition, with mixed results demonstrated in the literature. In one recent study that applied a comprehensive neuropsychological protocol to dementia and stroke‐free older population, a higher PVS burden was related to worse cognitive performance, specifically with information processing speed and executive function (EF), even after adjusting for demographic and common risk profiles. ¹⁴ Similarly, in a longitudinal study of Alzheimer's disease (AD) patients, PVS were associated with cognitive decline, independent of amyloid burden and other markers of small vessel disease. ¹⁵ Conversely, findings from a large memory clinic population that included patients with subjective cognitive decline, mild cognitive impairment (MCI), or dementia; ¹⁶ and, a meta‐analysis of five large population‐based studies from the UNIVRSE consortium, ¹⁷ found no association of PVS burden with cognitive dysfunction. Interestingly, despite the mixed results in the literature with PVS and cognition, white matter hyperintensities (WMH)—another commonly used MRI‐based marker of cerebral small vessel disease—are often associated with decreased information processing speed and EF, with minimal controversial findings in the literature. ¹⁸ , ¹⁹ , ²⁰

Of particular interest to glymphatic clearance and the perivascular space is the glial fibrillary acidic protein (GFAP), an intermediate filament protein found in astrocytes that is a commonly used marker of astroglial injury and glymphatic dysfunction. ²¹ , ²² Post‐mortem analyses of the deep white matter of patients with cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), an inherited form of cerebral small vessel disease, found increased numbers of GFAP‐positive clasmatodendritic astrocytes, particularly in the perivascular regions of the deep white matter—suggestive of severe compromise of the gliovascular unit. ²³ In AD patients, similar astrogliosis‐related increases in GFAP were observed in neuropathological analyses of WMH regions compared to normal‐appearing white matter. ²⁴

In contrast to imaging‐based biomarkers, plasma biomarkers have garnered recent interest in clinical research as less invasive and cost‐effective diagnostic tools for various neurodegenerative diseases causing dementia. ²⁵ Plasma GFAP has been associated with frontotemporal dementia (FTD), ²⁶ cerebral small vessel disease, ²⁷ AD, and vascular pathology. ²⁸ In vivo biomarker studies demonstrated associations of plasma GFAP with cognition and EF, where higher plasma GFAP levels were associated with lower EF scores among older adults with and without cognitive impairment. ²⁹ , ³⁰ Along with plasma neurofilament light chain (NfL), plasma GFAP is currently being considered as a non‐specific biomarker that will be included in a Revised Criteria for the Diagnosis and Staging of Alzheimer's Disease, by the Alzheimer's Association Workgroup. ³¹

Thus, given the cognitive outcomes potentially related to imaging‐ and plasma‐based biomarkers in cerebral small vessel disease and neurodegeneration, the main objective of this study was to examine the direct and indirect effects of PVS on speeded executive function (sEF) as mediated by GFAP and WMH, in a clinical neurodegenerative patient population.

2. METHODS

2.1. Participants

From the Ontario Neurodegenerative Disease Research Initiative (ONDRI) patient cohort, participants clinically diagnosed with Alzheimer's disease/mild cognitive impairment (AD/MCI), FTD, and cerebrovascular disease (CVD), with available clinical, demographic, MRI, neuropsychological, genetic and plasma biomarkers data were selected for analysis. ONDRI recruited participants from 14 tertiary care centers across Ontario, Canada, each of whom had passed preliminary screening. Detailed inclusion/exclusion criteria, demographics, and cohort characteristics of the ONDRI study were published previously. ³² , ³³ Briefly, the prevailing consensus‐based clinical diagnostic criteria at the time of enrollment for each ONDRI‐focused disease were implemented at each tertiary clinic: AD/MCI patients met National Institute on Aging—Alzheimer's Association criteria for dementia or MCI due to probable AD, with annual tracking of MCI patients to ensure accurate characterization of the cohort; FTD patients included possible or probable behavioral variants of frontotemporal degeneration, agrammatic/non‐fluent and semantic variants of primary progressive aphasia, and possible or probable progressive supranuclear palsy. The CVD cohort comprised individuals who experienced mild to moderate acute ischemic stroke, transient ischemic attack, or subcortical infarction, at least 3 months prior to enrollment, or severe white matter disease with subcortical infarcts, in compliance with the National Institute of Neurological Disorders and Stroke and the Canadian Stroke Network vascular cognitive impairment harmonization standards. Evidence of stroke required clinical imaging confirmation; however, individuals with large vessel occlusive infarction causing severe neurological deficits were excluded. Additionally, the AD/MCI patients’ MRI scans were assessed by a neuroradiologist to exclude individuals with non‐AD‐related causes for cognitive impairment. Ethics approval was obtained from the Research Ethics Board at each participating site. All participants provided written informed consent.

RESEARCH IN CONTEXT

1. Systematic review: MRI‐visible perivascular spaces (PVS) suggest glymphatic waste clearance dysfunction, with conflicting findings on cognition impact. Prior work links impaired glymphatic clearance to white matter injury and astrocytic reactivity in neurodegenerative diseases. White matter hyperintensities (WMH) and plasma glial fibrillary acidic protein (GFAP) associated with executive dysfunction. No in vivo studies explored PVS's indirect effects on speeded executive function (sEF) via WMH and plasma GFAP in neurodegenerative populations.

2. Interpretation: We demonstrated a small but significant effect of PVS volume on sEF, mediated by plasma GFAP and WMH.

3. Future directions: Our findings emphasize assessing these markers in memory clinics when faced with multiple co‐pathologies. They support a cascade: glymphatic dysfunction, perivascular enlargement, astroglial GFAP activation, neuroinflammatory cycle, and WMH eventually leads to sEF deficits. Future studies should focus on longitudinal data, include larger samples across different diseases, and explore other markers of glymphatic dysfunction and neuroinflammation.

2.2. Neuroimaging

The 3 Tesla MRI included the following sequences: three‐dimensional (3D) T1‐weighted (T1), interleaved proton density and T2‐weighted (T2), T2‐fluid attenuated inversion recovery (FLAIR), and T2*‐weighted gradient echo imaging sequences. Imaging protocol details were previously published and included in a Supplemental File. ³⁴ All MRI images were evaluated to ensure excellent imaging quality by a medical biophysicist (R.B.) and for clinical incidental findings by a licensed neuroradiologist (S.S.). Neuroimaging‐based markers for cerebral small vessel disease (WMH, PVS, lacunes, and infarcts), normal‐appearing gray matter (NAGM) volumes for gray matter (GM) atrophy, and total intracranial volume (TIV) for head‐size correction, were quantified using ONDRI's standardized image‐processing pipeline ³⁴ in compliance with the Standards for Reporting Vascular Changes on Neuroimaging (STRIVE) criteria. ¹³

In brief, volumetric data was acquired using the semi‐automated ONDRI post‐processing neuroimaging pipeline—a comprehensive, multi‐feature (T1, PD/T2, FLAIR) segmentation algorithm based on the SABRE‐LE (Semi‐automated brain extraction‐ Lesion explorer) and FLEX (Fuzzy lesion extraction) procedures ³⁵ , ³⁶ , ³⁷ and optimized for the ONDRI study. ³⁴ Each individual automated segmentation output mask was then manually checked and corrected as needed by neuroimaging analysts, who attained a DICE similarity index and intraclass correlation coefficient, both > 0.9. After the volumetric data was generated, the data was run through a novel multivariate outlier detection algorithm for data quality evaluation by an independent Neuroinformatics team ³⁸ and fed back to the Neuroimaging team for further manual checking (and correction if required), as an iterative quality control procedure.

2.3. Neuropsychological assessment

All patients underwent the ONDRI standardized neuropsychological test protocol which provided a comprehensive assessment of cognition and behavior. ³⁹ Screening was done for English comprehension, visual acuity, auditory acuity, and physical impairments which may confound test performance. The neuropsychological test scores were standardized (i.e., z‐transformed) using the mean and standard deviation of the normal participants from the Brain Eye Amyloid Memory Study (clinicaltrials.gov: NCT02524405). A composite measure for sEF was computed using mean z‐scores from the Digit Symbol Modalities Test, and the Trail Making Test Part B time, where increasing z‐scores were associated with increased speed, that is, better performance. The Trail Making Test Part B time was log‐transformed prior to standardization. Montreal Cognitive Assessment (MoCA) tests were also performed on all study participants.

2.4. Plasma biomarkers and genetics

Blood samples were drawn from all participants at baseline and shipped to LifeLabs (Etobicoke, Ontario, Canada) where they were immediately processed upon receipt for plasma isolation by centrifugation at 2000 × g rpm for 15 min at 4°C, and then stored at ‐80°C. They were then shipped on dry ice to the OBI Biobank Sample Reception at the Robarts Research Institute (Western University, London, ON, Canada), and stored there until shipment to the Clinical Neurochemistry Laboratory (University of Gothenburg, Mölndal, Sweden) for measurements. The concentrations of GFAP and NfL were measured using the Human Neurology 4‐plex E ultra‐sensitive Single molecule array (Simoa) immunoassay (Item 103670, Quanterix). The measurements were performed on an HD‐X Analyzer according to instructions from the manufacturer (Quanterix, Billerica, MA, USA). The measurements were performed in one round of experiments using one batch of reagents by board‐certified laboratory technicians who were blinded to clinical data. As a general index of directionality, higher GFAP and NfL are considered more pathological. Plasma NfL was included in the analysis as a relevant biomarker that is being considered in future clinical criteria. ³¹

Apolipoprotein E epsilon 4 (APOE E4) carrier statuses for each participant were mapped using the calls for the APOE risk alleles rs429358(CT) and rs7412(CT) extracted from the ONDRISeq custom next‐generation sequencing‐based panel data, as previously described. ⁴⁰ , ⁴¹

2.5. Statistics

To assess the effect of PVS volumes on sEF through GFAP level and WMH volume, a mediation analysis was applied using the PROCESS macro v3.5 in IBM SPSS (v28.0.1.1, SPSS Inc., Chicago, IL, USA). Bias‐corrected bootstrapping with 5000 replications and 95% confidence interval (CI) was performed to estimate the variability of direct, indirect, and total effects. We identified covariates by conducting separate Pearson correlations and linear regressions between demographic, medical history, and biomarker variables, on sEF and PVS. The following variables that were significant were selected as covariates in the main analyses: age, years of education, clinical diagnosis, hypertension, NAGM, NfL, lacunes, and presence of apolipoprotein E (APOE) E4. All variables were standardized with the mean and standard deviation of our sample. All neuroimaging variables were head‐size corrected via TIV; PVS and WMH were log‐transformed due to skewness.

3. RESULTS

The final sample size was 333: 155 patients with CVD, 126 patients with AD/MCI, and 52 patients with FTD. The mean age of our sample was 69.7 (± 7.7) years, the sample was mainly male (62.8%), highly educated (14.8 ± 2.9 years), with a mean total MoCA score of 23.7 ± 3.5 (Table 1). Although statistically significant, WMH and GFAP were only mildly correlated (Pearson r = 0.154, p = 0.005).

TABLE 1.

Demographics and neuroimaging volumetrics of the study population

No. of participants	N = 333
Demographics
Age, years	69.7 ± 7.7
Sex, male	209 (62.8%)
Level of education, years	14.8 ± 2.9
Montreal Cognitive Assessment, Total score	23.7 ± 3.5
Alcohol use (more than 7 drinks per week)	60 (18%)
Clinical diagnosis
Alzheimer's disease/Mild cognitive impairment	125 (37.8%)
Cerebrovascular disease	155 (46.5%)
Frontotemporal dementia	52 (15.6%)
Vascular risk factors
Hypertension	166 (49.8%)
Dyslipidemia	206 (61.9%)
Diabetes	67 (20.1%)
Currently smoking	22 (6.6%)
Waist/hip circumference	0.93± 0.09
Coronary artery disease	34 (10.2%)
Atrial fibrillation	38 (11.4%)
Neuroimaging
Total intracranial volume, mean (SD), cm3	1232.01 ± 136.92
Gray matter volume, mean (SD), cm3	533.18 ± 52.53
MRI‐visible perivascular spaces volume, median (IQR), mm3	43.0 (56.50)
WMH, median (IQR), cm3	3.54 (6.22)
Plasma biomarkers
GFAP, pg/mL	118 ± 99.44
Neurofilament light chain, pg/mL	28.43 ± 24.37

Abbreviations: GFAP, glial fibrillary acidic protein; IQR, interquartile range; WHM, white matter hyperintensities.

PVS was significantly associated with sEF (c = ‐0.125 ± 0.054, 95% bootstrap [CI] [‐0.2309, ‐0.0189], p = 0.021, black on Figure 1). This effect was mediated by both GFAP and WMH. Specifically, the indirect effect of PVS on sEF via GFAP was significant (a₁b₁ = ‐0.0197 ± 0.0108, 95% CI [‐0.0442, ‐0.0021], a₁: p = 0.005; b₁: p = 0.029, blue on Figure 1); the indirect effect via WMH was also significant (a₂b₂ = ‐0.0316 ± 0.0180, 95% CI [‐0.0712, ‐0.0016], a₂: p < 0.001; b₂: p = 0.031, green on Figure 1). The total indirect effect of PVS on sEF via GFAP and WMH was significant (‐0.051 ± 0.022, 95% CI [‐0.097; ‐0.013], red in Figure 1).

FIGURE 1.

Plasma GFAP and WMH volume mediate the effect of PVS on sEF. Mediation Model showing plasma glial fibrillary acidic protein and white matter hyperintensity volumes are significantly mediating the associations of perivascular space volumes with speeded executive function. GFAP, glial fibrillary acidic protein; PVS, perivascular spaces; SDMT, Digit Symbol Modalities Test; sEF, speeded executive function; TMTb, Trail Making Test Part B; WMH, white matter hyperintensities (created with BioRender.com)

4. DISCUSSION

In this unique clinical cohort of mixed neurodegenerative and cerebrovascular patients, this study's findings shed light on the complex in vivo biomarker relationships between PVS, plasma GFAP, WMH, and cognition. Specifically, we demonstrated that PVS has a small but significant effect on sEF and that this effect was mediated via plasma GFAP and WMH in this clinical population. Our in vivo findings support the increasing neuropathological evidence that perivascular enlargement from glymphatic clearance dysfunction, commonly observed in mixed cerebral small vessels and neurodegenerative disease, may be associated with increased GFAP expression in the perivascular regions. ²¹ , ²³ , ²⁴ , ⁴² The cascade of glymphatic clearance dysfunction, perivascular enlargement, astroglial GFAP activation, the neuroinflammatory cycle, and WMH (likely due to perivascular edema from venous collagenosis ⁴³ , ⁴⁴ ), eventually leads to downstream deficits in sEF observed in our neurodegenerative and CVD patients.

Previously regarded as benign radiological anomalies, PVS has garnered recent interest due to advances in our understanding of cerebral fluid circulation and glymphatic waste clearance. PVS, previously referred to as Virchow‐Robin spaces, are anatomically defined as the fluid‐filled channels surrounding the brain's vasculature, bound by the adventitia of the vessel and the astrocytic end‐feet. ⁹ Facilitated by aquaporin‐4 (AQP4) water channels expressed by astrocytes, glymphatic perivascular clearance of harmful metabolic waste, such as amyloid β, has been shown in preclinical models to be active primarily during deep sleep. ⁴ Indeed, the clearance of amyloid β, tau, and α‐synuclein, hallmark neuropathological protein aggregates in AD and other neurodegenerative diseases, has been a therapeutic target since their discovery over 100 years ago. Additionally, the role of perivascular neuroinflammation ⁴⁵ indicated by the activation of microglia and astrocytes, ⁴⁶ and the GFAP expression observed in WMH regions from cerebral small vessel disease, ²¹ , ²³ , ⁴² likely contributes to the overall impairment of sEF that was indicated by the PVS burden observed in our mixed neurodegenerative and CVD patients’ MRI. The mediating roles of GFAP and WMH on PVS may be the pieces to the puzzle that explain the mixed findings in the literature in regard to PVS and cognition.

Although grounded primarily from preclinical and neuropathological research findings, ample evidence regarding the clinical relevance of PVS burden is evident in the recent update to the Boston criteria. ⁴⁷ Used for in vivo clinical diagnosis of cerebral amyloid angiopathy (CAA), the Boston criteria v2.0 includes the assessment of T2‐weighted MRI for severe PVS burden—defined as > 20 PVS in the centrum semiovale of one hemisphere. As this criteria would mark the first clinical diagnostic criteria that include PVS burden, it is likely that further advances that include plasma GFAP in diagnostic criteria may soon follow. ³¹

There are several important unanswered factors that are highly relevant to this discussion that should be given attention in future work: (1) the inclusion of a sleep quality measure, given its role in glymphatic system activation; (2) the measurement of AQP4 as driver for the movement of fluids into perivascular channels; (3) a measure of amyloid β, tau, and/or α‐synuclein burden in our patient population; and 4) the role of APOE4 and pericyte degeneration in blood‐brain barrier dysfunction. ⁴⁸ , ⁴⁹ In addition to these unanswered factors, it is also important to note that we did not examine longitudinal changes in cognition which prevented us from inferring directionality in our models; and, although the overall sample was decently powered, the individual patient cohorts were relatively small, which limited our statistical ability to examine individual disease cohorts. However, our heterogeneous real‐world representation of patients attending memory clinics may be considered a strength of our findings. Finally, our plasma biomarker assay did not include plasma placental growth factor, which has recently been recognized as a relevant biomarker for vascular cognitive impairment. ⁵⁰

In clinical practice, our study findings emphasize the importance of T1, T2‐weighted, and T2‐FLAIR MRI for the assessment of WMH and PVS burden, and the acquisition of plasma biomarkers such as NfL and GFAP, when faced with many of the complex neurodegenerative and neurovascular diseases that cause dementia.

AUTHOR CONTRIBUTIONS

Conception of the research project: Allison A. Dilliott, Robert A. Hegele, Paula M. McLaughlin, Carmela M. Tartaglia, Ekaterina Rogaeva, Richard H. Swartz, Dar Dowlatshahi, Gustavo Saposnik, Malcolm A. Binns, Robert Bartha, Mario Masellis, and Sandra E. Black. Organization of the research project: Allison A. Dilliott, Robert A. Hegele, Paula M. McLaughlin, Courtney Berezuk, Richard H. Swartz, Gustavo Saposnik, Malcolm A. Binns, Morris Freedman, Robert Bartha, Mario Masellis, Sandra E. Black, and Joel Ramirez. Execution of the research project: Allison A. Dilliott, Robert A. Hegele, Paula M. McLaughlin, Christopher J.M. Scott, Courtney Berezuk, Miracle Ozzoude, Carmela M. Tartaglia, Ekaterina Rogaeva, David F. Tang‐Wai, Richard H. Swartz, Leanne Casaubon, Sanjeev Kumar, Dar Dowlatshahi, Jennifer Mandzia, Demetrios Sahlas, Corinne Fischer, Michael Borrie, Elizabeth Finger, Andrew Frank, Robert Bartha, Sean Symons, Henrik Zetterberg, Mario Masellis, Sandra E. Black, and Joel Ramirez. Data Curation of the research project: Allison A. Dilliott, Robert A. Hegele, Fuqiang Gao, Paula M. McLaughlin, Jennifer S. Rabin, Madeline E. Wood, Christopher J.M. Scott, Vanessa Yhap, Courtney Berezuk, Miracle Ozzoude, Walter Swardfager, Robert Bartha, Sean Symons, Henrik Zetterberg, Mario Masellis, Sandra E. Black, and Joel Ramirez. Supervision and resources of the research project: Robert A. Hegele, Carmela M. Tartaglia, Ekaterina Rogaeva, Richard H. Swartz, Leanne Casaubon, Dar Dowlatshahi, Jennifer Mandzia, Demetrios Sahlas, Gustavo Saposnik, Corinne Fischer, Michael Borrie, Ayman Hassan, Malcolm A. Binns, Morris Freedman, Elizabeth Finger, Andrew Frank, Robert Bartha, Sean Symons, Howard Chertkow, Mario Masellis, Sandra E. Black, and Joel Ramirez. Design of the statistical analysis: Daniela Andriuta, Julie Ottoy, and Joel Ramirez. Execution of the statistical analysis: Daniela Andriuta. Review and critique of the statistical analysis: Julie Ottoy, Myuri Ruthirakuhan, Malcolm A. Binns, and Joel Ramirez. Writing of the first draft of the manuscript: Daniela Andriuta. Review and critique of the manuscript: Julie Ottoy, Myuri Ruthirakuhan, Ginelle Feliciano, Allison A. Dilliott, Robert A. Hegele, Fuqiang Gao, Paula M. McLaughlin, Jennifer S. Rabin, Madeline E. Wood, Christopher J.M. Scott, Courtney Berezuk, Miracle Ozzoude, Walter Swardfager, Carmela M. Tartaglia, Ekaterina Rogaeva, David F. Tang‐Wai, Richard H. Swartz, Leanne Casaubon, Sanjeev Kumar, Dar Dowlatshahi, Jennifer Mandzia, Demetrios Sahlas, Gustavo Saposnik, Corinne Fischer, Michael Borrie, Ayman Hassan, Malcolm A. Binns, Morris Freedman, Elizabeth Finger, Andrew Frank, Robert Bartha, Sean Symons, Henrik Zetterberg, Howard Chertkow, Mario Masellis, Sandra E. Black, and Joel Ramirez. All co‐authors agree with the conditions noted on the Authorship Agreement Form.

CONFLICT OF INTEREST STATEMENT

D.A. reports no disclosures relevant to the manuscript. J.O. reports no disclosures relevant to the manuscript. M.R. reports no disclosures relevant to the manuscript. G.F. reports no disclosures relevant to the manuscript. A.A.D. reports no disclosures relevant to the manuscript. R.A.H. reports no disclosures relevant to the manuscript. F.G. reports no disclosures relevant to the manuscript. P.M.M. reports no disclosures relevant to the manuscript. J.S.R. reports no disclosures relevant to the manuscript. M.W.A. reports no disclosures relevant to the manuscript. C.J.M.S. reports no disclosures relevant to the manuscript. V.Y. reports no disclosures relevant to the manuscript. C.B. reports no disclosures relevant to the manuscript. M.O. reports no disclosures relevant to the manuscript. W.S. reports no disclosures relevant to the manuscript. J.Z. reports no disclosures relevant to the manuscript. M.C.T. reports no disclosures relevant to the manuscript. E.R. reports no disclosures relevant to the manuscript. D.F.T.‐W. reports no disclosures relevant to the manuscript. L.C. reports no disclosures relevant to the manuscript. S.K. reports no disclosures relevant to the manuscript. D.D. reports no disclosures relevant to the manuscript. J.M. reports no disclosures relevant to the manuscript. D.S. reports no disclosures relevant to the manuscript. G.S. reports no disclosures relevant to the manuscript. C.E.F. reports no disclosures relevant to the manuscript. M.B. reports no disclosures relevant to the manuscript. A.H. reports no disclosures relevant to the manuscript. M.A.B. reports no disclosures relevant to the manuscript. M.F. reports no disclosures relevant to the manuscript. H.C. reports no disclosures relevant to the manuscript. E.F. reports no disclosures relevant to the manuscript. A.F. reports no disclosures relevant to the manuscript. R.B. reports no disclosures relevant to the manuscript. S.S. reports no disclosures relevant to the manuscript. H.Z. reports no disclosures relevant to the manuscript. R.H.S. reports no disclosures relevant to the manuscript. M.M. reports no disclosures relevant to the manuscript. S.E.B. reports no disclosures relevant to the manuscript. J.R. reports no disclosures relevant to the manuscript. D.A. reports funding (travel and meetings) from Biogen, Roche, Teva, Novartis, Bristol‐Myers Squibb, Genzyme, and Sanofi outside the submitted work. H.Z. has served at scientific advisory boards and/or as a consultant for Abbvie, Acumen, Alector, Alzinova, ALZPath, Annexon, Apellis, Artery Therapeutics, AZTherapies, Cognito Therapeutics, CogRx, Denali, Eisai, Merry Life, Nervgen, Novo Nordisk, Optoceutics, Passage Bio, Pinteon Therapeutics, Prothena, Red Abbey Labs, reMYND, Roche, Samumed, Siemens Healthineers, Triplet Therapeutics, and Wave, has given lectures in symposia sponsored by Alzecure, Biogen, Cellectricon, Fujirebio, Lilly, and Roche, and is a co‐founder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program (outside submitted work). C.E.F. receives grant funding from NIA, NIH, CCNA, CIHR, ADDF, TDRA, Mito2i, the Hilary and Galen Weston Foundation, and Novo Nordisk. Author disclosures are available in the supporting information.

Supporting information

Supporting Information

Supporting Information

Andriuta D, Ottoy J, Ruthirakuhan M, et al. Perivascular spaces, plasma GFAP, and speeded executive function in neurodegenerative diseases. Alzheimer's Dement. 2024;20:5800–5808. 10.1002/alz.14081

DATA AVAILABILITY STATEMENT

The ONDRI data are publicly available via the Ontario Brain Institute (https://braininstitute.ca/).