Evaluation of an in vivo biomarker of arteriolosclerosis (ARTS) and its associations with cognition and multimodal ATN(V) biomarkers in a cardiometabolic‐risk enriched community cohort

Abstract

INTRODUCTION

We evaluated associations between an in vivo (magnetic resonance imaging [MRI]) marker of arteriolosclerosis (ARTS) and multimodal neuroimaging and plasma ATN(V) biomarkers.

METHODS

Among 238 participants with both amyloid and tau positron emission tomography (PET) scans within 1 year of MRI, we examined multivariable adjusted models relating ARTS with structural MRI (cortical thickness/volume, white matter hyperintensities [WMH]), diffusion MRI (fractional anisotropy [FA], mean diffusivity [MD], NODDI free water [FW]), cerebral blood flow, plasma biomarkers (p‐tau217, Aβ42/40, neurofilament light, glial fibrillary acidic protein [GFAP]), and PET imaging.

RESULTS

As expected, ARTS was most strongly linked to age and greater WMH burden and diffusion‐based indices of microstructural disruption (FA, MD, and FW). ARTS was elevated in ATN biomarker‐positive groups (highest in A+T+N+) and was associated with greater neurodegeneration and higher plasma biomarker levels, GFAP in particular.

DISCUSSION

ARTS relates to other markers of vascular brain injury, neurodegeneration, amyloid and tau pathology within the ATN(V) framework, and inflammation.

Highlights

• arteriolosclerosis (ARTS) scores are elevated in ATN‐positive individuals, most prominently in A+T+N+

• ARTS may exert stage‐specific effects on neurodegeneration rather than track A/T burden.

• Strong ARTS‐glial fibrillary acidic protein (GFAP) association suggests role of astroglial activation in vascular‐related neurodegeneration

• Findings underscore the importance of vascular contributions to Alzheimer's disease and related dementias (AD/ADRD) pathophysiology, especially in high‐cardiometabolic‐risk populations

1. BACKGROUND

Alzheimer's disease and related dementias (AD/ADRD) are increasingly recognized as multifactorial disorders, in which vascular injury often plays a role alongside amyloid and tau pathology. Approximately 70%–80% of dementia (DEM) cases present with mixed etiology, the majority exhibiting concomitant vascular pathology that often emerges in midlife. ¹ , ² The revised ATN(V) framework for AD, encompassing amyloid (A), tau (T), neurodegeneration (N), and vascular injury (V), provides a unifying structure for integrating these processes into a single biomarker model. Cerebral small‐vessel disease (cSVD) frequently co‐occurs with AD/ADRD pathology ³ , ⁴ , ⁵ and is increasingly recognized as a contributor to AD/ADRD. ⁶ , ⁷

cSVD represents various forms of vascular pathology visible on brain magnetic resonance imaging (MRI) ⁸ as well as microstructural changes not visible on imaging; therefore, multiple (V) biomarkers likely need to be evaluated within the context of the revised ATN(V) framework. Arteriolosclerosis, hardening and loss of elasticity of arterioles or small arteries, is a hallmark of cSVD frequently associated with neurodegeneration, impaired cognitive functioning ⁹ and cognitive decline, ¹⁰ and increased odds of AD. ¹¹ Tau and amyloid pathology may potentiate the microinfarct burden that characterizes cSVD, ⁴ highlighting a synergistic interaction between vascular injury and neurodegenerative processes. Despite growing post‐mortem evidence suggesting associations of arteriolosclerosis with AD pathology, in vivo measurement remains challenging. ¹² Traditional MRI markers of cSVD are nonspecific, indirect proxies that lack specificity for underlying arteriolosclerosis pathology, particularly when considered in isolation.

To advance in vivo measurement of arteriolosclerosis and increase understanding of the vascular contributions to cognitive impairment and dementia (VCID), the ARTS marker was developed ¹² and tested within the MarkVCID consortium, a machine learning‐based tool trained on ex vivo MRI and neuropathology data and validated across three independent cohorts. The ARTS marker offers a semi‐automated resource for indexing arteriolosclerosis in vivo across the research community. While ARTS represents a major methodological advancement, its relationship to AD/ADRD biomarkers, particularly within the ATN(V) framework, remains underexplored. Two prior studies reported that ARTS was associated with peripheral vascular risk factors, small‑vessel injury on MRI, and cognitive outcomes. ¹³ , ¹⁴ However, studies have not examined ARTS within the ATN(V) framework using amyloid/tau positron emission tomography (PET) or plasma biomarkers such as phosphorylated tau 217 (p‑tau217), neurofilament light chain (NfL), or glial fibrillary acidic protein (GFAP), limiting insight into the biological mechanisms underlying these associations.

RESEARCH IN CONTEXT

1. Systematic review: We used PubMed and conference proceedings to identify studies linking cerebral small‑vessel disease, arteriolosclerosis (ARTS), and ATN(V) biomarkers. Prior studies show that small‑vessel disease contributes to cognitive decline, but no work has evaluated magnetic resonance imaging (MRI)‑derived ARTS within the ATN(V) framework, including amyloid and tau positron emission tomography (PET) or plasma biomarkers, limiting insight into how arteriolosclerosis relates to Alzheimer's disease and related dementias (AD/ADRD).

2. Interpretation: Higher ARTS scores were associated with greater white matter hyperintensity (WMH) burden, microstructural disruption, neurodegeneration, and elevated plasma biomarkers, particularly glial fibrillary acidic protein (GFAP). ARTS was highest in A+T+N+ individuals, suggesting that microvascular injury intersects with amyloid, tau, and inflammatory pathways and may exert stage‑specific neurodegenerative effects.

3. Future directions: As an indirect marker of arteriolosclerosis, ARTS should be related to cerebrovascular markers (microbleeds, lacunes, enlarged perivascular spaces), tested for sensitivity to vascular risk modification on ARTS trajectories and cognition, evaluated against large‑vessel atherosclerosis to capture distinct small‑vessel mechanisms, and studied in relation to vascular and metabolic comorbidities in AD/ADRD.

To address this gap, we investigated ARTS in relation to multimodal imaging (primary) and plasma (secondary) biomarkers within the ATN(V) framework ¹⁵ and established cSVD markers in a deeply phenotyped, community‐based cohort of older adults enrolled in the Clinical Core of the Wake Forest Alzheimer's Disease Research Center (WF ADRC). As described previously, the WF ADRC cohort is characterized by elevated rates of obesity, hypertension (HTN), and cardiovascular disease relative to national averages, ¹⁶ making it well‐suited for establishing relationships between ARTS and AD/ADRD pathology in the context of cardiometabolic disorders. All included participants had MRI, amyloid, and tau PET; a subset of these participants had plasma biomarkers. Tau PET imaging, in particular, a key component of the ATN framework, ¹⁵ offers a unique opportunity to contextualize vascular contributions within the broader landscape of AD/ADRD biomarkers given its established role as a robust predictor of clinical progression and its ability to stratify individuals by risk. ¹⁷ , ¹⁸ , ¹⁹ We hypothesized that higher ARTS scores would be associated with greater amyloid and tau burden and altered plasma biomarker profiles, reflecting the vascular contributions to AD/ADRD pathology within the ATN(V) framework. To our knowledge, this is the first study to integrate ARTS with amyloid/tau PET and plasma biomarkers within the ATN(V) framework in a high‐cardiometabolic‑risk, community‐based cohort.

2. METHODS

2.1. Participants

Participants between the ages of 55 and 85 were recruited from the surrounding community into the WF ADRC via the WF ADRC Clinical Core supported by efforts of the Outreach, Engagement & Recruitment Core. ²⁰ , ²¹ No participants received or were recruited with the intent of receiving or being recommended for disease‐modifying treatments or therapies, nor were participants referred to our study for treatment (previously published exclusionary criteria are provided in the supplement for reference). Participants underwent a standard evaluation, including the National Alzheimer's Coordinating Center (NACC) protocol for clinical research data collection, clinical exams, neurocognitive testing, neuroimaging (MRI, amyloid PET, and tau PET), blood collection for plasma‐based biomarkers, and genotyping for the apolipoprotein E (APOE) ε4 allele. WF ADRC participants (n = 238) who received cognitive diagnoses at adjudication (see cognitive status) by June 1, 2025, and had an MRI and both amyloid and tau PET within 1 year (see Figure S1 for greater details) were selected for the ARTS ATN sub‐study. The Wake Forest Institutional Review Board approved all activities as described; written informed consent was obtained for all participants and/or their legally authorized representatives.

2.2. Clinical and cognitive evaluation

An expert panel of investigators including neuropsychologists, neurologists, and geriatricians provided adjudication of cognitive status in accordance with current National Institute of Aging‐Alzheimer's Association guidelines. Following the determination of clinical cognitive diagnosis for cognitively unimpaired (CU), mild cognitive impairment (MCI), ²² and DEM, ²³ adjudication of cause/type of cognitive impairment or DEM was assessed using neuroimaging and fluid biomarkers. Cognition was assessed using a modified Preclinical Alzheimer's Cognitive Composite (mPACC5) comprised of scores obtained from in‐person visits on the MMSE, Free and Cued Selective Reminding Test (FCSRT), Craft Story delayed verbatim recall, Digit Symbol Substitution Test, and category fluency assessments. ²⁴ , ²⁵

2.3. HTN and impaired glucose tolerance

Brachial blood pressure was measured in a seated position after a 5‐min rest in a quiet room as described previously ²⁰ and was categorized according to 2017 ACC/AHA guidelines, ²⁶ with HTN status defined as systolic blood pressure (SBP) ≥ 140 mmHg and/or having a history HTN or current use of antihypertensive medications as indicated on UDSv3 form d1. Fasting blood glucose levels were measured via an oral glucose tolerance test (OGTT) from serial blood draws; impaired glucose tolerance (IGT) was defined as fasting glucose tolerance > 100 mg/dL or hemoglobin A1c > 5.6%. ¹⁶ Cardiometabolic index (CMI) was calculated using the product of two ratios (waist/height × triglycerides/high density lipoprotein [HDL]). ¹⁶ , ²⁷ Additional details on medications can be found in the Supporting Information.

2.4. Imaging biomarkers

2.4.1. MRI acquisition and processing

Participants were scanned on a research‐dedicated 3‐Tesla Siemens Skyra MRI (32‐channel head coil) scanner. Detailed image acquisition parameters have been previously published. ²⁰ , ²¹ , ²⁸ T1, T2 FLAIR, and DTI/NODDI scans were acquired. T1 image processing included SPM12 (www.fil.ion.ucl.ac.uk/spm) CAT12 normalization and tissue segmentation, as well as regional volume and cortical thickness and total intracranial volume (ICV) estimation using FreeSurfer v7.2 (https://surfer.nmr.mgh.harvard.edu). FreeSurfer bilateral and unilateral (left and right) total cortical gray matter (GM) volume, bilateral hippocampal volume, and temporal lobe cortical thickness (bilateral entorhinal, inferior/middle temporal, and fusiform) were calculated. White matter hyperintensity (WMH) volume (WMH; lesions of presumed ischemic origin) was segmented by the lesion growth algorithm (LGA) implemented in the LST toolbox v2.0.15, running in SPM12 using FLAIR and T1 images. ²⁹ A log transformation was applied to WMH to account for its skewed distribution. ³⁰ , ³¹ DTI and NODDI processing details have been described previously. ²⁰ , ²¹ Briefly, NODDI free water (FW) was averaged in global supratentorial GM, global supratentorial white matter (WM), and bilateral hippocampal GM; DTI fractional anisotropy (FA) and mean diffusivity (MD) were averaged in global supratentorial WM. The Johns Hopkins University (JHU) DTI atlas ³² and AAL atlas ³³ , ³⁴ were overlaid on template‐space parameter images (FW, FA, or MD) that were tissue‐specific (WM or GM) to extract mean signal across all supratentorial WM tracts and supratentorial GM regions of interest (ROIs), respectively. The same process was used to calculate mean global GM and WM cerebral blood flow (CBF) from Arterial Spin labeling scans using a multiphase pseudo‐continuous arterial spin labeling (MP‐PCASL) sequence as published in greater detail elsewhere. ²⁰ MRI‑derived vascular biomarkers included WMH volume, global WM DTI metrics (FA, MD), NODDI FW in GM and WM, and global GM and WM CBF. MRI‑derived vascular biomarkers selected to represent the V component of the ATN(V) framework, capturing complementary features of small‑vessel disease pathology, included WMH volume, global WM DTI metrics (FA, MD), NODDI FW in GM and WM, and global GM and WM CBF.

2.4.2. Aβ‐PET and tau‐PET imaging

Fibrillar Aβ brain deposition on PET was assessed with [¹¹C]‐Pittsburgh Compound B (PiB). ³⁵ Tau PET with [18F]Flortaucipir (FTP) was used for assessing tau neurofibrillary tangle deposition. ³⁶ , ³⁷ PET imaging details have been described in detail elsewhere ²¹ , ³⁸ and have been provided in the Supporting Information for convenience.

2.4.3. Biomarker classification

In the current analyses, biomarker classification (e.g., ATN) was based exclusively on neuroimaging data. Amyloid (PET; [A]), tau (PET; [T]), and neurodegeneration (MRI; [N]) positivity (A+/T+/N+) were defined and evaluated using a combination of visual reads and both well‐defined and sample‐specific a priori thresholds. ³⁹ , ⁴⁰ , ⁴¹ Visual and quantitative methods were used to fit separate models assessing associations between ARTS and biomarker positivity. Visual reads were conducted (by M.M.R., S.N.L., and J.R.B.) in accordance with the prescribing information for Tauvid (FTP; Eli Lilly and Company, 2020). Sample‐derived thresholds were validated using a combination of Gaussian‐mixture modeling and receiver operating characteristic curves using published methods. ¹⁶ , ²¹ , ⁴² For primary analyses assessing ARTS within the ATN classification framework, amyloid‐PET positivity was determined via visual read (corresponding to an approximately global standardized uptake value ratio [SUVr] ≥ 1.21 ¹⁶ ). Tau‐PET positivity was defined by a meta‐temporal SUVr ≥ 1.21. ⁴³ Our primary index of MRI‐based neurodegeneration positivity (N+) was defined using hippocampal volume, adjusted for ICV (e.g., HCV ≤ 0.454). ⁴⁴ , ⁴⁵ We also explored defining N+ based on a combination of ICV‐adjusted global and hippocampal GM volumes and cortical thickness. ⁴⁴ , ⁴⁵ There was high agreement between methods for classifying N+ (91% of cases were N+ across approaches) and mean ARTS scores were nearly identical across the two classification schemes (mean ARTS: HCV‐only N+: = −0.309, N− = 0.393; Combined N+ = −0.294, N− = −0.400); thus, this method is not discussed further. Plasma AD/ADRD biomarkers, described further below, were included in confirmatory and exploratory analyses.

2.5. Plasma biomarkers

A subset of the sample with MRI and amyloid and tau PET had plasma biomarker data (p‐tau217, n = 184; Aβ42/Aβ40, NfL, and GFAP, n = 166). Plasma AD/ADRD biomarkers were collected from participants after a minimum 8‐h (water only) fast. Blood was processed within 30 min of collection as described previously. ¹⁶ , ²¹ Batch shipments were sent to the National Centralized Repository for Alzheimer's Disease and Related Dementias (NCRAD) Biomarker Assay Laboratory for analysis. Plasma Aβ42/40, NfL, and GFAP were assessed using the Quanterix Simoa Neurology 4‐Plex E and p‐tau181 v2 Advantage Kits on a Quanterix Simoa HD‐X. ²¹ Plasma p‐tau217 samples, collected from 2017 to 2023, were processed in duplicate using ALZpath Simoa p‐tau 217 v2 assay kits on a Quanterix HD‐X at Neurocode (Bellingham, WA). ¹⁶ P‐tau217 analyses were conducted on 1st thawed samples with the same platform and assay lot number across samples. Duplicates with CVs > 20% (maximum upper limit) ⁴⁶ , ⁴⁷ , ⁴⁸ or missing one value were repeated. Kit QC controls were run with each plate. The coefficient of variation (CV; mean = 4.47; standard deviation [SD] = 3.64) for p‐tau217 was well within previously established reference limits and similar to other cohorts, as recently published. ⁴⁹ All plasma biomarker concentrations were quantified in pg/mL, consistent with standard reporting for Simoa‑based assays. P‐tau217 served as our primary plasma AD‐specific biomarker in the current study. ¹⁶ , ⁴⁹ , ⁵⁰ We also evaluated the p‐tau217/Aβ42 ratio. ⁵¹

2.6. ARTS score calculation

ARTS is a supervised machine learning‐based classifier of brain arteriolosclerosis. ARTS was trained on ex‐vivo neuropathological data with demonstrated good performance predicting in‐vivo brain arteriolosclerosis. ¹² , ¹³ , ⁵² Raw T1‐weighted and FLAIR images, together with TOPUP/Eddy‐corrected diffusion MRI NIfTI volumes, were submitted to the standard ARTS pipeline. ARTS then performed its native preprocessing and analysis workflow to derive global WMH burden and tract‐based FA, which were used as predictors in the ARTS classifier along with participant age at time of scan and self‐reported sex (see https://markvcid.partners.org/markvcid1‐protocols‐resources; https://markvcid.partners.org/sites/default/files/markvcid2/protocols/ARTS_protocol_v7.13.23.pdf). ¹² Diffusion data were provided as TOPUP/Eddy‐corrected NIfTI volumes, thus the diffusion‐related preprocessing steps were skipped when running ARTS; all orientation information was preserved, enabling appropriate registration to the ARTS template. ARTS outputs a single score per scan session; higher ARTS scores indicate a higher likelihood of arteriolosclerosis.

2.7. Statistical analysis

Participant demographics were compared across cognitive status groups and by ATN classification using chi‐square tests and one‐way analysis of variance. Associations between biomarkers were evaluated using Spearman's rank correlation, appropriate for non‑parametric and potentially non‑linear relationships consistent with our prior work. ¹⁶ Multivariable general linear models (GLM) examined relationships between neuroimaging and plasma variables with ARTS in unadjusted (Model 1; see Table 1), and adjusted models controlling for age, sex, race, education, APOE‐ε4 carrier status, and days between PET acquisition or blood‐sample collection and MRI scans (Model 2). Plasma biomarker models additionally controlled for body mass index (BMI) and kidney function (estimated glomerular filtration rate [eGFR]) as these factors can influence plasma biomarker levels due to physiological reasons (i.e., renal function or blood volume). ²¹ , ⁵⁰ , ⁵³ , ⁵⁴ Gamma regression models were also fit when comparing ARTS and PET biomarkers given the severity of skewness in amyloid‐PET and tau‐PET estimates. In post‐hoc analyses, we examined if associations between ARTS and AD/ADRD biomarkers were modified by cognitive status, sex, race, and APOE‐ε4 status. Significance level was set at an alpha of α = 0.05. Multiple comparisons corrections were performed across all statistical tests assessing association between ARTS and MRI, PET, and plasma AD/ADRD biomarkers using the Benjamini–Hochberg false‐discovery rate (q). ⁵⁵ All statistical tests were conducted in R (RStudio Team, 2020).

TABLE 1.

Participant characteristics.

		Overall	CU	MCI	DEM
Parameter	N	N = 238 a	n = 132 a	n = 80 a	n = 26 a	p‐value b
Age (MRI)	238	71 (8)	71 (8)	72 (7)	70 (8)	0.200
Sex:female	238	150 (63%)	95 (72%)	41 (51%)	14 (54%)	0.006
Race	238					0.200
White		185 (78%)	99 (75%)	62 (78%)	24 (92%)
AA/Black		50 (21%)	32 (24%)	16 (20%)	2 (7.7%)
AI/AN		1 (0.4%)	0 (0%)	1 (1.3%)	0 (0%)
NH/PI		1 (0.4%)	0 (0%)	1 (1.3%)	0 (0%)
Asian		1 (0.4%)	1 (0.8%)	0 (0%)	0 (0%)
Education	238	15.91 (2.57)	16.08 (2.40)	15.61 (2.70)	15.96 (3.03)	0.500
mPACC5	236	−0.59 (1.13)	0.06 (0.67)	−1.07 (0.88)	−2.51 (0.77)	<0.001
Smoker	238	76 (32%)	38 (29%)	27 (34%)	11 (42%)	0.400
APOE‐ε4+	229	74 (32%)	34 (27%)	29 (38%)	11 (44%)	0.110
ARTS	238	−0.37 (0.17)	−0.38 (0.15)	−0.35 (0.18)	−0.33 (0.16)	0.200
SYSBP	238	135 (21)	133 (22)	138 (20)	137 (20)	0.150
HTN	238	159 (67%)	79 (60%)	63 (79%)	17 (65%)	0.016
LDL	251	102.98 (31.28)	103.32 (31.40)	99.99 (31.68)	110.69 (28.99)	0.300
HDL	251	63.73 (19.69)	63.94 (19.98)	62.77 (17.78)	65.65 (24.11)	>0.90
OGTT (BSL)	255	97.24 (16.62)	96.56 (16.16)	99.71 (18.38)	93.22 (12.06)	0.300
IGT	238	141 (59%)	77 (58%)	52 (65%)	12 (46%)	0.400
BMI	238	27.8 (6.4)	28.8 (6.4)	26.9 (6.9)	25.5 (3.7)	0.009
eGFR	230	78 (15)	77 (14)	78 (16)	81 (16)	0.200
HX TBI	238	6 (2.5%)	3 (2.3%)	2 (2.5%)	1 (3.8%)	0.800
HX sleep apnea	238	67 (28%)	34 (26%)	27 (34%)	6 (23%)	0.400
PET imaging
Aβ+ (visual read)	238	102 (43%)	39 (30%)	43 (54%)	20 (77%)	<0.001
Aβ+ (CL > 24)	233	85 (36%)	30 (23%)	37 (47%)	18 (69%)	<0.001
Aβ‐PET (SUVr)	234	1.40 (0.45)	1.26 (0.34)	1.50 (0.48)	1.79 (0.57)	<0.001
Aβ‐PET (centiloids)	233	28 (44)	14 (31)	38 (47)	67 (56)	<0.001
Aβ‐PET MRI INT	238	24 (222)	19 (235)	13 (231)	79 (62)	0.900
TAU‐PET+ (SUVr)	238	97 (41%)	33 (25%)	49 (61%)	15 (58%)	<0.001
TAU PET (SUVr)	238	1.27 (0.28)	1.17 (0.08)	1.31 (0.22)	1.63 (0.55)	<0.001
TAU‐PET MRI INT	238	86 (107)	85 (117)	81 (96)	102 (86)	0.900
Neuroimaging
GMV	238	0.68 (0.04)	0.69 (0.04)	0.67 (0.04)	0.65 (0.04)	<0.001
HCV	238	0.48 (0.08)	0.51 (0.07)	0.46 (0.07)	0.40 (0.08)	<0.001
CORTTHICK	238	2.75 (0.15)	2.78 (0.13)	2.75 (0.15)	2.58 (0.18)	<0.001
WMH (log/ICV)	238	−1.69 (1.26)	−1.91 (1.23)	−1.45 (1.34)	−1.42 (0.94)	0.025
DTI FA	238	0.512 (0.020)	0.513 (0.020)	0.510 (0.021)	0.509 (0.019)	0.500
DTI MD	238	0.82 (0.05)	0.80 (0.04)	0.82 (0.05)	0.84 (0.04)	<0.001
NODDI FW	237	0.173 (0.022)	0.170 (0.022)	0.173 (0.020)	0.187 (0.024)	0.004
ASL WM CBF	234	17.7 (3.3)	18.2 (3.5)	17.1 (2.9)	16.8 (3.0)	0.039
ASL GM CBF	234	36 (8)	37 (8)	34 (7)	31 (6)	<0.001
Plasma
p‐tau217	184	0.45 (0.36)	0.36 (0.26)	0.57 (0.39)	0.66 (0.58)	0.003
p‐tau217/Aβ42	158	0.08 (0.07)	0.06 (0.05)	0.11 (0.09)	0.12 (0.11)	0.008
Aβ42/40	166	0.053 (0.010)	0.053 (0.010)	0.053 (0.010)	0.048 (0.008)	0.018
NfL	166	16 (7)	14 (6)	18 (9)	21 (8)	<0.001
GFAP	166	136 (68)	127 (63)	138 (60)	179 (95)	0.038

Abbreviations: Aβ, amyloid beta; APOE, apolipoprotein E; ARTS, arteriolosclerosis; BMI, body mass index; CBF GM, cerebral blood flow in gray matter; CBF WM, cerebral blood flow in white matter; CL, centiloids; CORTTHICK, cortical thickness; CU, cognitively unimpaired; DEM, dementia; DTI FA, diffusion tensor imaging fractional anisotropy; DTI MD, diffusion tensor imaging mean diffusivity; eGFR, estimated glomerular filtration rate; GFAP, glial acid fibrillary acid protein; GMV, total gray matter brain volume adjusted for intracranial volume; HCV, hippocampal volume adjusted for intracranial volume; HDL, high‐density lipoprotein; HTN, hypertension; HX TBI, history of traumatic brain injury; IGT, impaired glucose tolerance; INT, interval (days between MRI and PET scan acquisition); LDL, low‐density lipoprotein; MCI, mild cognitive impairment; NfL, neurofilament light chain; NODDI FW, neurite orientation diffusion index free water; OGTT, oral glucose tolerance test; SUVr, standardized uptake volume ratio; SYSBP, systolic blood pressure; WMH, log‐transformed white matter hyperintensity volume adjusted for intracranial volume.

^a Mean (SD); n (%).

^b Kruskal‐Wallis rank sum test; Pearson's Chi‐squared test; Fisher's exact test.

3. RESULTS

3.1. Participant characteristics

A total of 238 participants completed brain MRI and both PiB (amyloid) PET (43% A+) and FTP (tau) PET (41% T+) within 1 year of MRI, and all underwent clinical and cognitive evaluation (Table 1). The mean age of participants was approximately 71 years, 21% were Black, 63% were female, and 32% were APOE‐ε4 carriers (APOE‐ε4+). Regarding cardiometabolic health, 67% had HTN, and 32% with a history of smoking. Approximately 63% of the analytical sample reported current or recent use hypertensive medications (CU = 59% [n = 78]; MCI = 70% [n = 56]; DEM = 54% [n = 14]; ∼84% of those classified as having HTN (as described above). Irrespective of HTN status, 76% of those with high blood pressure, as defined here, reported hypertensive medication use (18% with uncontrolled blood pressure).

Demographics were comparable to and representative of the larger WF ADRC cohort (see Table S1). Participants were adjudicated by NIA‐AA criteria as having normal cognition (n = 132 [55%]), MCI (n = 80 [34%]), or DEM (n = 26 [11%]). CU had a higher proportion of females compared to MCI and DEM groups. MCI and DEM participants had higher blood pressure compared to CU, while MCI had the highest proportion of participants with HTN (71%; Table 1). Neuroimaging and plasma biomarkers differed by cognitive status as previously reported in the larger WF ADRC cohort ¹⁶ , ²¹ and were altered in the impaired groups (CU < MCI < DEM) for neuroimaging measures, except for DTI FA (p > 0.05), and each of the plasma biomarkers assessed, including p‐tau217, Aβ42/40, NfL, and GFAP (all p < 0.05).

3.2. ARTS scores

3.2.1. General descriptives

ARTS scores ranged from −0.75 to 0.10 (mean ± SD = −0.37 ± 0.17), suggesting a lower overall likelihood of arteriolosclerosis in our cohort compared to the reference samples, which were slightly older than our ADRC cohort (see Sections 2 and 4). ARTS scores were higher on average in females, participants with HTN, and participants with a history of smoking (all p < 0.05; see Table 1 and Figure S2). ARTS scores did not differ by cognitive status, race, APOE‐ε4 carriership, impaired glucose status, or hypertensive medication use (independent of hypertensive status), or a history of TBI or sleep apnea (all p > 0.05). Consistent with ARTS as an aging vascular biomarker, ARTS scores were strongly positively associated with age (rho = 0.62, p < 0.001; Figure S3). ARTS was also negatively correlated with mPACC5 scores in unadjusted models (rho = 0.17, p = 0.011). Finally, ARTS scores were not associated with lipids indexed by low‑ and high‑density lipoprotein cholesterol, nor with cardiometabolic status (e.g., CMI; all p > 0.05). Raw unadjusted Spearman correlation matrices are presented in Figure S4.

3.3. Biomarkers

3.3.1. Associations of ARTS with amyloid and tau PET deposition

Higher ARTS score was significantly associated with greater levels of amyloid deposition and greater tau PET deposition in unadjusted models (Table 2, Model 1; Figure 1; also see Figure S5), and in models adjusted by sex, race, education, APOE‐ε4 carriership, and the interval in days between PET and MRI (Model 2). Results were nearly identical using gamma regression (see Table S2). Associations of ARTS with PET positivity of A+ and T+ are presented in the Supporting Information (also see Table S3).

TABLE 2.

Association of ARTS with amyloid and tau PET deposition.

	Model 1 (unadjusted)						Model 2 (adjusted)
Parameter	n	R 2	β	SE	p	q	n	R 2	β	SE	p	q
Aβ (SUVr)	234	0.07	0.70	0.17	<0.001	<0.001	225	0.22	0.54	0.22	0.013	0.017
Aβ (CL)	233	0.06	63.26	16.73	<0.001	<0.001	225	0.20	56.15	21.16	0.009	0.011
TAU (SUVr)	238	0.02	0.24	0.10	0.020	0.024	229	0.10	0.41	0.14	0.003	0.005

Note: All models were fit using robust standard errors. Model 1 was unadjusted. Model 2 was adjusted for age, sex, race, education, APOE, and the interval between PET and MRI scan acquisition. Covariates significantly contributing to model performance are summarized in Table S4.

Abbreviations: Aβ, amyloid beta; APOE, apolipoprotein E; ARTS, arteriolosclerosis; CL, Centiloids; FDR, false discovery rate; MRI, magnetic resonance imaging; PET, positron emission tomography; q, FDR‐corrected p‐values; R ², coefficient of determination (% variance explained); SE, standard error; SUVr, standard‐uptake value ratio.

FIGURE 1.

Forest plot of adjusted models evaluating associations between ARTS and AD/ADRD biomarkers. Forest plot of standardized beta‐weights and their respective 95% confidence intervals from adjusted models testing associations between ARTS and AD/ADRD biomarkers. Unadjusted scatterplots showing associations between ARTS and neuroimaging and plasma biomarkers are provided in Figures S6 and S7, respectively. Aβ, amyloid beta; AD/ADRD, Alzheimer's disease and related dementias; ARTS, arteriolosclerosis; CBF GM, cerebral blood flow in gray matter; CBF WM, cerebral blood flow in white matter; CL, Centiloids; CORTTHICK, cortical thickness; DTI FA, diffusion tensor imaging fractional anisotropy; DTI MD, diffusion tensor imaging mean diffusivity; GFAP, glial acid fibrillary acid protein; GMV, total gray matter brain volume adjusted for intracranial volume; HCV, hippocampal volume adjusted for intracranial volume; NfL, neurofilament light chain; NODDI FW, neurite orientation diffusion index free water; SUVr, standardized uptake volume ratio; WMH, log‐transformed white matter hyperintensity volume adjusted for intracranial volume.

3.3.2. Associations of ARTS with MRI measures

Higher ARTS was significantly associated with poorer overall brain health, including reduced GM volume, DTI FA, and CBF in GM, and higher WMH volume and NODDI FW across all models assessed (Table 3; Figure 1). ARTS was not significantly associated with CBF in WM before or after adjustment (all p > 0.05; Table 3; also see Table S4 and Figure S5).

TABLE 3.

Associations between ARTS and MRI measures.

A. Structural
	Model 1 (unadjusted)						Model 2 (adjusted)
	n	R 2	β	SE	p	q	n	R 2	β	SE	p	q
GMV	238	0.21	−0.12	0.02	<0.001	<0.001	229	0.43	−0.10	0.02	<0.001	<0.001
HCV	238	0.10	−0.14	0.03	<0.001	<0.001	229	0.36	−0.11	0.03	0.001	0.001
CORTTHICK	238	0.05	−0.20	0.06	0.001	0.002	229	0.07	−0.21	0.01	0.008	0.011
WMH	238	0.48	5.24	0.36	<0.001	<0.001	229	0.52	4.96	0.44	<0.001	<0.001
DTI FA	238	0.45	−0.08	0.01	<0.001	<0.001	229	0.61	−0.11	0.01	<0.001	<0.001
DTI MD	238	0.63	0.22	0.01	<0.001	<0.001	229	0.72	0.25	0.01	<0.001	<0.001
NODDI FW	237	0.26	0.07	0.01	<0.001	<0.001	228	0.34	0.06	0.01	<0.001	<0.001

B. Functional
	Model 1 (unadjusted)						Model 2 (adjusted)
	n	R 2	β	SE	p	q	n	R 2	β	SE	p	q
CBF WM	234	0.00	−0.70	1.32	0.598	0.616	225	0.19	−1.14	1.60	0.475	0.505
CBF GM	234	0.03	−8.38	3.04	0.006	0.009	225	0.31	−9.81	3.50	0.005	0.008

Note: Associations between ARTS and MRI measures before and after adjustment. ARTS was most strongly associated with WMH and DTI metrics as expected (see Section 2; also see Section 4 for caveats). Covariates significantly contributing to model performance are summarized in Table S4. Model 1, unadjusted; Model 2, adjusted for age, sex, race, education, APOE.

Abbreviations: APOE, apolipoprotein E; ARTS, arteriolosclerosis; CBF, cerebral blood flow; CORTTHICK, temporal meta‐region of interest cortical thickness; DTI, diffusion tensor imaging; FA, fractional anisotropy; FW, freewater; GM, gray matter; GMV, total gray matter volume adjusted for intracranial volume; HCV, hippocampal volume adjusted for intracranial volume; MD, medial diffusivity; MRI, magnetic resonance imaging; NODDI, neurite orientation dispersion index; q, FDR‐corrected p‐values; R ², coefficient of determination (% variance explained); SE, standard error; WM, white matter; WMH, log‐transformed white matter hyperintensity volume adjusted for intracranial volume.

3.3.3. Associations of ARTS with plasma AD/ADRD biomarkers

Higher ARTS scores were significantly associated with elevated plasma biomarker levels in unadjusted models (Model 1; Table 4, Figure 1) and models adjusted for age, sex, race, education, APOE genotype, BMI, eGFR, and the time interval between blood draw and MRI acquisition (Model 2; Table 4, Figure 1; also see Table S4 and Figure S6).

TABLE 4.

Associations between ARTS and plasma AD/ADRD biomarker levels.

	Model 1 (unadjusted)						Model 2 (adjusted)
Parameter	n	R 2	β	SE	p	q	n	R 2	β	SE	p	q
p‐tau217	184	0.06	0.52	0.16	0.001	0.002	180	0.24	0.47	0.19	0.015	0.018
p‐tau217/Aβ42	158	0.06	0.10	0.03	0.002	0.003	155	0.23	0.11	0.04	0.007	0.010
Aβ42/40	166	0.03	−0.01	0.00	0.026	0.030	160	0.15	−0.00	0.01	0.672	0.672
NfL	166	0.10	14.01	3.22	<0.001	<0.001	160	0.36	6.34	3.37	0.062	0.068
GFAP	166	0.20	176.65	27.71	<0.001	<0.001	160	0.34	107.65	34.83	0.002	0.004

Note: All models were fit using robust standard errors. Model 1 was unadjusted. Model 2 was adjusted for age, sex, race, education, APOE, BMI, eGFR, and the interval between blood collection and MRI scan acquisition. Covariates significantly contributing to model performance are summarized in Table S4.

Abbreviations: AD/ADRD, Alzheimer's disease and related dementias; APOE, apolipoprotein E; ARTS, arteriolosclerosis; BMI, body mass index; eGFR, estimated glomerular filtration rate; FDR, false discovery rate; GFAP, glial fibrillary acidic protein; MRI, magnetic resonance imaging; NfL, neurofilament light chain; q, FDR‐corrected p‐values; R ², coefficient of determination (% variance explained); SE, standard error.

3.4. ARTS and ATN framework

ARTS scores were elevated in biomarker positivity groups (A+T−N− [n = 25], A+T+N− [n = 27], A+T+N+ [40]) as compared to A−T−N− (n = 85; all p < 0.05). ARTS scores showed a small stepwise increase across ATN groups, consistent with a monotonic trend (β = 0.043 per step, p < 0.001). Pairwise contrasts confirmed that the largest difference was between biomarker‑negative (A−T−N−) and fully biomarker‑positive (A+T+N+) participants, with intermediate groups showing smaller or nonsignificant differences (Figure 2). ARTS scores tended to be higher on average in N+ groups, including those not in the canonical ATN grouping framework for AD (e.g., AD/ADRD; A−T+N−, A−T−N+, A−T+N+, A+T−N+; see Tables S5 and S6 for ARTS and cohort descriptives across both canonical and all ATN categorizations, respectively).

FIGURE 2.

ARTS scores across diagnostic and biomarker groups. ARTS scores are shown stratified by (A) cognitive status (left panel) and (B) ATN biomarker classification (right panel). Statistics and effect sizes for comparisons are shown in panel C. ARTS scores did not differ appreciably by clinical diagnosis. In contrast, higher (worse) ARTS scores were observed in participants with increasing ATN positivity compared to biomarker‑negative individuals. In the right panel, OTHER represents cases not represented by the canonical AD‐specific ATN framework (e.g., A−T+N− [n = 23], A−T−N+ [n = 20], A−T+N+ [n = 9], A+T−N+ [n = 12]). These individuals were not included in stratification analyses (though see Table S6 for a detailed breakdown of descriptives for all biomarker positivity groups as ARTS score tended to be higher in N+ groups). A, amyloid; ARTS, arteriolosclerosis; CI‐H, upper bounds of 95% confidence interval; CI‐L, lower bounds of 95% confidence interval; EF, effect size; MAG, effect size magnitude; N, neurodegeneration; N, sample size; T, tau.

3.5. Post‐hoc analyses

Associations between ARTS scores and neuroimaging and plasma AD/ADRD biomarkers were largely independent of demographic, clinical, and health factors, including hypertensive status. However, in adjusted models, we observed several outcome‐dependent modifications by diagnostic status, sex, and race (all interaction p < 0.05; see Figure S7). First, higher ARTS scores were more strongly associated with elevated amyloid PET levels in DEM as compared to CU and MCI participants. Second, associations between higher ARTS and lower hippocampal volume and cortical thickness estimates were stronger in female compared to male participants. Third, associations between higher ARTS scores and greater WMH burden were stronger in Black compared to White participants. No other interactions were observed (data not shown).

4. DISCUSSION

This targeted study provides novel evidence linking ARTS, a multimodal, machine learning‐based index of arteriolosclerosis, to neuroimaging and plasma biomarkers of AD/ADRD within the ATN(V) framework in a deeply phenotyped, community‐based cohort. Higher ARTS scores were associated with poorer brain health, including reduced global and hippocampal brain volume, lower cortical thickness, and diminished GM CBF. These structural findings are consistent with prior work demonstrating that vascular (e.g., structural) pathology contributes to neurodegeneration and cognitive decline and relates to AD pathology. ⁹ , ⁵⁶ , ⁵⁷ , ⁵⁸ In addition, higher ARTS scores were associated with putative biomarkers of V pathology (greater WMH burden and lower WM microstructural integrity) as well as with plasma biomarkers of neurodegeneration (higher plasma levels of neurofilament light) and neuroinflammation (indirectly indexed by GFAP). Relative to markers of neurodegeneration, associations with amyloid and tau PET and with plasma p‐tau217 were relatively modest and were attenuated after adjusting for cognitive status, given that these observed relationships were strongest among individuals with DEM. Finally, ARTS scores were elevated in neurodegeneration positive (N+) participants and were found to increase across the stages of the AD biomarker continuum (e.g., A+/T+/N+). By embedding ARTS within the ATN(V) framework and integrating tau/amyloid PET with plasma and perfusion metrics, this study extends prior ARTS work from structural cSVD correlates to biomarker‑defined AD pathways, revealing stage‑dependent and risk‑modified associations.

We focused our analyses on a subset of the WF ADRC with ATN(V) biomarkers. We were particularly interested in tau PET because tau pathology is more closely linked to clinical symptoms and disease progression. ¹⁸ , ¹⁹ , ³⁷ , ⁵⁹ , ⁶⁰ , ⁶¹ , ⁶² Impaired clearance of amyloid‐β due to vascular dysfunction may accelerate amyloid accumulation ⁶³ and contribute to tau progression. ⁶⁴ Evidence to date has been inconsistent regarding associations between large‑vessel arteriosclerosis (e.g., arterial stiffness, pressure pulsatility) and AD/ADRD biomarkers. One study reported links with tau but not amyloid, localized to cortical regions that accumulate tau early in disease. ⁶⁵ Another found no association with tau PET burden but did observe arteriosclerosis‑related amyloid deposition, modified by antihypertensive medication use, highlighting the potential importance of vascular co‑pathology in AD clinical trials. ⁶⁶ Such inconsistencies likely reflect differences in cohort characteristics, the focus on large‑artery disease alone, and the examination of individual etiologies in mostly unimpaired participants, where tau typically emerges later in the disease course.

This work extends prior literature with a more comprehensive view of ADRD pathology by incorporating biomarkers of arteriosclerosis and relating them to imaging and plasma measures of ATN(V). By integrating PET and plasma biomarkers, our findings extend Lamar et al., ¹³ who showed ARTS predicted cognitive decline in older African Americans, and Fleischman et al., ¹⁴ who linked ARTS to incident MCI, DEM, and stroke (see Section 4.1). Our findings suggest that vascular pathology, captured by ARTS (e.g., small vessel arteriolosclerosis), may interact with tau and amyloid deposition in a stage‐dependent or risk‐modified manner. These patterns are consistent with the idea that vascular dysfunction (e.g., cSVD) may increase the burden of AD/ADRD‐related pathology, and that such relationships may be especially relevant for individuals with midlife vascular risk factors (e.g., HTN). Because WMH and diffusion tensor imaging‐derived FA are the primary imaging inputs to ARTS, strong associations with global summary metrics (e.g., WMH, DTI FA) were expected. However, it is worth noting that WMH and FA measures captured in the WF ADRC are not identical to those used in ARTS (see Sections 2 and 4.1).

Notably, although our study is not designed to support direct comparisons between previous ARTS‐related studies, we did observe that ARTS scores in our cohort were lower than those reported in the initial validation by Makkinejad et al., despite ∼67% of participants having HTN, which is a rate comparable to other evaluated cohorts and the broader WF ADRC. ¹² , ¹⁴ , ⁵² Although speculative, this discrepancy may reflect our cohort's relatively younger age, on average, and the high prevalence of antihypertensive medication use in the WF ADRC, both of which could attenuate ARTS scores. Recently, Arfanakis et al. reported follow‐up validation demonstrating high internal consistency and replication of ARTS scores across several cohorts independent of Makkinejad et al. ¹² , ⁵² Although these validation studies support the robustness of ARTS scores derived across different cohorts and scanner parameters, additional follow‐up in the larger WF sample and other representative cohorts is needed to determine whether the lower ARTS scores in our cohort reflect scanner‐ or acquisition‐related factors (e.g., number of diffusion directions) versus true cohort differences. Likewise, given that 94% of hypertensive individuals in our sample were taking antihypertensive medications, we were unable to distinguish controlled from uncontrolled HTN.

Overall, ARTS scores showed small stepwise increases across ATN groups; as expected, the largest pairwise difference was observed when comparing A−T−N− to A+T+N+. ARTS scores were modestly associated with core imaging Alzheimer's disease biomarkers and were most strongly correlated with MRI‐based indicators of small‐vessel disease, particularly WMH and FA (again, see Section 4.1), and vascular risk factors. ARTS was elevated in participants with HTN and a history of smoking, consistent with prior work. ¹⁴ , ⁶⁷ , ⁶⁸ , ⁶⁹ , ⁷⁰ Small‐vessel injury linked to HTN is well established to be associated with increased WMH and FA abnormalities ⁶⁷ , ⁶⁸ , ⁶⁹ , ⁷⁰ and may also contribute to inflammation, as indirectly indexed by GFAP in our study. Notably, in addition to expected associations with WMH and FA, we observed relationships with NODDI Freewater and GM CBF, suggesting that ARTS may capture broader microvascular and perfusion‐related injury. We also did not observe associations between ARTS and lipid measures (e.g., HDL, low‐density lipoprotein [LDL]) or composite cardiometabolic indices. Lipid profiles may be more strongly linked to large‐vessel atherosclerosis, ⁷¹ , ⁷² whereas associations between cSVD and cardiometabolic burden (e.g., obesity, diabetes) may be mediated by HTN or inflammation rather than directly influencing ARTS. ⁷³ , ⁷⁴ Additional work is needed to further understand the role of inflammation (I) in the context of ATN(V) and the mechanisms (e.g., endothelial dysfunction, impaired glymphatic clearance of amyloid, or neurovascular coupling) that may underlie ARTS–AD/ADRD relationships. It remains unclear whether ARTS reflects a distinct pathophysiological signature or merely recapitulates known associations between vascular risk factors and cSVD‐related neuroimaging metrics. Importantly, we fully acknowledge that ARTS was trained to predict ex‐vivo arteriosclerosis, where it has been shown to outperform other established biomarkers including in vivo estimates of WMH burden. ¹² , ⁵²

4.1. Limitations and future directions

Our findings should be interpreted in light of limitations and technical considerations. First, the sample was restricted to participants with tau PET, reducing statistical power but enabling targeted analysis of downstream AD pathology. Although characteristics of the reduced sample were similar to the larger WF ADRC cohort, as a single‐site study, replication in larger more representative samples is needed. Although our findings suggest associations between ARTS and AD/ADRD were largely independent of demographic factors, several outcomes (e.g., sex and APOE) warrant further investigation. Likewise, given the prevalence of hypertensive medication use, we were not able to assess the impact of controlled versus uncontrolled HTN. Second, observed associations between ARTS and amyloid PET and tau PET, and plasma p‐tau217 were relatively modest in this subset of participants. Further validation of novel relationships between ARTS and AD biomarkers and consideration of cases that do not fall into the typical ATN schema are needed. Third, the cross‑sectional nature of this study precludes causal inference. Longitudinal studies are required to determine whether ARTS predicts tau accumulation, amyloid progression, or cognitive decline. Fourth, ARTS is an indirect and surrogate biomarker of arteriolosclerosis derived from MRI features (WMH and FA) ¹² , ⁵² and may not capture the full spectrum of cSVD. Future work should examine its relationship to other cerebrovascular markers such as cerebral microbleeds, lacunes, and enlarged perivascular spaces. ⁷⁵ , ⁷⁶ , ⁷⁷ , ⁷⁸ Moreover, vascular and metabolic comorbidities, including HTN, diabetes, obesity, and renal dysfunction, are common in aging populations and may interact with genetic and demographic risk factors to confer greater risk of AD/ADRD. ⁷⁹ , ⁸⁰ , ⁸¹ , ⁸² Advanced analytic approaches that integrate distinct imaging and clinical features could help disentangle how arteriolosclerosis specifically contributes to cognitive decline and DEM risk. Finally, peripheral and cerebral measures of arteriosclerosis can diverge, as compensatory mechanisms in the brain may partially protect cerebral vessels from systemic arterial stiffening. ⁸³ This distinction underscores the importance of examining both large‑ and small‑vessel disease, which likely act through complementary but distinct pathways. In future work, we plan to directly test how microvascular injury interacts with systemic vascular changes to impact AD/ADRD pathology.

5. CONCLUSION

To our knowledge, this is the first study to integrate the MRI‑derived ARTS score with established neuroimaging and plasma biomarkers of AD/ADRD in the ATN(V) framework. Higher ARTS scores were consistently associated with poorer brain health, including reduced volumes, lower cortical thickness, diminished GM CBF, and elevated plasma markers of neuroaxonal injury and neuroinflammation. Novel associations with tau PET extend prior work and situate vascular injury alongside amyloid, tau, and neurodegeneration in the ATN(V) continuum. These findings highlight the importance of addressing vascular health in midlife and support ARTS as a non‐invasive, scalable biomarker of small‑vessel disease. More broadly, they reinforce the hypothesis that arteriolosclerosis contributes to AD/ADRD pathology through converging vascular and neurodegenerative pathways. Future studies should determine whether modifying vascular risk alters ARTS trajectories, downstream neurodegenerative processes, and cognitive outcomes, and should test whether ARTS diverges from large‑vessel atherosclerosis, commonly influenced by lipid metabolism, by capturing distinct small‑vessel mechanisms such as impaired clearance, neurovascular coupling deficits, and neuroinflammation.

AUTHOR CONTRIBUTIONS

All authors made substantial contributions and contributed to the final draft. MDR, TMH, and SNL contributed to the conception and design. MDR performed data analyses. All authors approved the version to be published and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

ETHICS STATEMENT

The Wake Forest Institutional Review Board approved all activities as described.

CONSENT

Written informed consent was obtained for all participants and/or their legally authorized representatives.

CONFLICT OF INTEREST STATEMENT

Marc Rudolph, Melissa Rundle, Richard Barcus, Kathryn Alphin, Kiran Solingapuram Sai, Thomas Register, Shannon Risacher, and Timothy Hughes have no conflicts of interest to disclose. Samuel Lockhart is a full‐time employee of Perceptive, Inc. James Bateman has received funding from the Alzheimer's Association and honoraria from Efficient CME, PeerView CME, and Novo Nordisck in the last three years. Michelle Mielke consults for or serves on advisory boards for Biogen, Eisai, Lilly, Merck, Roche, and Siemens Healthineers. Suzanne Craft reports disclosures for vTv Therapeutics, T3D Therapeutics, Cyclerion Inc., and Cognito Inc. Author disclosures are available in the supporting information.

Supporting information

Supplementary Material

Supplementary Material: alz71483‐sup‐0002‐SuppMat.docx