Temporal dynamics of white matter hyperintensities related to Alzheimer's disease in adults with Down syndrome

Abstract

INTRODUCTION

White matter hyperintensities (WMH) are common in Down syndrome (DS), yet their longitudinal evolution and associations with Alzheimer's disease (AD) remain unclear.

METHODS

Longitudinal MRI study, including 80 DS adults and 53 euploid controls. WMH were segmented on serial FLAIR using a longitudinal pipeline. We assessed the effects of demographic, genetic factors, AD clinical stage, AD‐related fluid, and cerebrovascular biomarkers on annual WMH volume changes.

RESULTS

In DS, annual WMH changes were relatively stable until age 40, and then exhibited fluctuations, with a significant decrease at the group level. Declines were larger in symptomatic cases, particularly in periventricular and fronto‐parieto‐occipital regions. Higher baseline WMH and microbleeds presence related to greater WMH reduction. Visual ratings and adjustment for white matter volume supported the robustness of the results.

DISCUSSION

WMH trajectories were heterogeneous in DS and declined over time with AD symptoms. This unexpected reduction may reflect different underlying pathological processes, including neurodegeneration or neuroinflammation.

Highlights

• The evolution of white matter hyperintensities (WMH) in Down syndrome (DS) is variable over time and with age.

• After age 40, more cases experienced significant WMH variations, primarily decreases.

• In DS, annual WMH change decreases with Alzheimer's symptoms.

• Higher baseline WMH and microbleeds are related to greater annual WMH reduction.

• WMH reduction likely reflects distinct processes, for example, degeneration or inflammation.

1. BACKGROUND

Down syndrome (DS), caused by a full or partial trisomy of chromosome 21, is the most common genetic form of intellectual disability (ID), affecting approximately 6 million people worldwide. ¹ Advances in healthcare have significantly increased the life expectancy of individuals with DS to around 60 years, ² which has led to an increased prevalence of age‐related conditions, notably Alzheimer's disease (AD). Indeed, AD is the leading cause of death in adults with DS. ³ This heightened risk is mainly due to the triplication of the amyloid precursor protein (APP) gene on chromosome 21, which is both necessary and sufficient to cause early‐onset AD neuropathology in DS. ⁴ Amyloid‐β (Aβ) accumulation begins during adolescence, and by the age of 40, nearly all individuals with DS exhibit the neuropathological features characteristic of AD. ⁵ , ⁶ In addition, Aβ can aggregate within the walls of leptomeningeal and cortical blood vessels, leading to the development of cerebral amyloid angiopathy (CAA). ⁷

Interestingly, despite the low prevalence of hypertension and arteriolosclerosis in DS, ⁸ , ⁹ neuroimaging studies reveal changes classically associated with small vessel disease (SVD) in adults with DS. ¹⁰ , ¹¹ Hence, white matter hyperintensities (WMH), a key neuroimaging hallmark of SVD, increase more markedly with age and begin at an earlier age in individuals with DS than in the euploid population. ¹⁰ , ¹² , ¹³ Two large studies in DS have reported that WMH start to increase around age 40 and are strongly associated with the clinical progression of AD and AD pathology in DS. ¹¹ , ¹² A recent longitudinal study from the Alzheimer's Biomarker Consortium‐Down Syndrome (ABC‐DS) showed that WMH trajectories are heterogeneous, with declines over time in cognitively stable adults with DS and increases, at the group level, with advancing AD stages, even though a subset of symptomatic participants exhibited decreases. These findings suggest that WMH fluctuates over time in DS and may reflect pathophysiological processes related to the overproduction of the Aβ and AD.

With the introduction of anti‐amyloid monoclonal antibodies into clinical practice as well as the development of new therapeutic agents, it is essential to delineate the natural progression of cerebrovascular lesions, particularly WMH, in populations likely to benefit from, such as individuals genetically predisposed to AD, including those with DS. ¹⁴ WMH can resemble certain secondary effects of these drugs, such as amyloid‐related imaging abnormalities (ARIA), complicating the interpretation of treatment‐related imaging changes. Furthermore, WMH may be considered as primary or secondary outcomes in clinical trials targeting vascular integrity or neuroinflammatory pathways, underscoring the need for a thorough understanding of their natural history and pathophysiology to accurately assess therapeutic efficacy.

The aim of this study is to investigate the longitudinal changes of WMH in DS and their associations with demographic, clinical, and AD biomarker data, as well as other vascular markers. Based on previous evidence, we hypothesize that WMH accumulation accelerates during the symptomatic stages and is associated with neurodegeneration.

2. METHODS

2.1. Study design and participants

This single‐center, longitudinal cohort study recruited adults with DS from the population‐based Down‐Alzheimer Barcelona Neuroimaging Initiative (DABNI) cohort ¹⁵ and euploid, cognitively unimpaired control individuals (HC) from the Sant Pau Initiative on Neurodegeneration (SPIN) cohort. ¹⁶ Participants of both sexes were aged 18 or older, with at least two magnetic resonance imaging (MRI) scans and comprehensive neurological and neuropsychological assessments. The study was approved by the Sant Pau Ethics Committee following the standards for medical research in humans recommended by the Declaration of Helsinki. All participants or their legally authorized representatives gave written informed consent before enrolment.

RESEARCH IN CONTEXT

1. Systematic review: White matter hyperintensities (WMH) are common in adults with Down syndrome (DS), a genetically determined form of Alzheimer's disease (AD). The prevalence and severity of WMH in DS increase markedly with age, particularly after ∼40 years, coinciding with near‐universal AD neuropathology in DS. WMH also relates to AD clinical progression and pathology, suggesting an interplay between small vessel disease (SVD) and AD. However, the longitudinal evolution of WMH in DS and their association with AD pathology remain unexplored.

2. Interpretation: Annual WMH changes fluctuated over time in DS. During the 20s‐30s, most individuals exhibited minimal WMH changes; however, after age 40, many showed notable variations, mainly decreases. Greater reductions of annual WMH changes were observed in individuals with AD symptoms, higher baseline WMH burden, and the presence of microbleeds. Declines in WMH, not fully explained by white matter atrophy, suggest additional pathophysiological mechanisms.

3. Future directions: Future research combining long‐term and multimodal neuroimaging with post mortem validation could provide a better characterization of the WMH etiology.

2.2. Clinical assessment

Adults with DS underwent a specific neuropsychological assessment, including the Cambridge Cognitive Examination for Older Adults with Down Syndrome (CAMCOG‐DS) Spanish version ¹⁷ to assess the global cognition by evaluating orientation, language, memory, attention, praxis, abstract thinking, and perception. Additionally, ID was categorized as mild, moderate, or severe/profound based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, ¹⁸ and caregivers’ report of the individual's best‐ever level of functioning and the score of the Kaufman Brief Intelligence Test Spanish version. ¹⁹

Participants were clinically classified in a consensus meeting between neurologists and neuropsychologists, after independent visits and blinded to biomarker data, into the following four groups: (1) asymptomatic (aDS, i.e., no clinical or neuropsychological suspicion of AD‐related cognitive decline), (2) prodromal AD (pDS, i.e., evidence of cognitive decline due to AD, but no significant impact on baseline activities of daily living), (3) AD dementia (dDS, i.e., when the cognitive decline impaired the daily activities) and (4) uncertain (uDS, i.e., when there were medical, pharmacological, or psychiatric condition interfering with cognition or daily activities, but no suspicion of neurodegenerative origin). To maximize statistical power, we combined in this study pDS and dDS into a single symptomatic group (sDS). Clinical evaluations were performed at baseline and follow‐up visits. Based on changes in clinical diagnosis along the AD continuum, participants were further categorized as progressors (aDS progressing to sDS at follow‐up), or stable (no change or uDS at any visit).

The HC group consisted of cognitively unimpaired euploid individuals who had no cognitive complaints, scored 0 on the Clinical Dementia Rating scale and/or 1 in the Global Deterioration Scale–Functional Assessment Staging, and performed within the normal range on neuropsychological evaluations regarding their age and education. None reported any neurological or psychiatric disorders or other major medical illnesses. ¹⁶

2.3. Neuroimaging data

2.3.1. Image acquisition

All participants underwent at least two MRI sessions at Hospital del Mar (3T Philips‐Achieva) or Hospital Clinic (3T Siemens Prisma), separated by a minimum interval of 6 months. MRI protocols include high‐resolution three‐dimensional T1‐weighted (T1w), fluid‐attenuated inversion recovery (FLAIR), and susceptibility weighted images (SWI). Only participants scanned with the same MRI protocol at baseline and follow‐up were included. Detailed acquisition parameters of the imaging protocols are provided in Table S1 (see Supporting Information).

2.3.2. WMH quantification

2.3.2.1. T1 preprocessing

Structural T1w images were preprocessed with the Computational Anatomy Toolbox 12 version 8 (CAT12v8) ²⁰ toolbox of Statistical Parametric Mapping 12 (SPM12, Welcome Centre for Human Neuroimaging, University College London, Queen Square Institute of Neurology). For each participant and time point, CAT12 was used to segment total gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) maps. Hammer's atlas was also coregistered into the native T1w space at each visit.

2.3.2.2. WMH longitudinal segmentation

To accurately segment WMH across multiple time points, we employed the longitudinal pipeline implemented in the Lesion Segmentation Toolbox version 3.0.0 (LST) ²¹ for SPM12. This pipeline is specifically designed to compare lesion probability maps previously generated by the Lesion Prediction Algorithm (LPA) ²² across multiple time points.

As a preprocessing step, FLAIR scans were coregistered into their corresponding T1w images using Advanced Normalization Tools (ANTs). ²³ We then applied the LPA method to each time point to generate lesion probability maps of WMH based on a logistic regression classifier that labels each voxel as a lesion or non‐lesion. The resulting probability maps were binarized using a threshold of 0.3, following our previous work. ¹²

Next, the WMH longitudinal pipeline of LST performed rigid between‐time‐point coregistration of each participant's FLAIR images. The resulting space transforms were subsequently applied to the lesion probability maps, ensuring that for each participant, all FLAIR images and corresponding lesion maps were aligned in a common space.

Finally, relative changes in intensity among FLAIR scans were computed for all voxels that were labeled as a lesion in at least one time point. Through an iterative procedure, the pipeline distinguishes true lesion change from natural variability in FLAIR signal across time points. This yielded refined and binarized lesion maps for each time point that were jointly informed by the full longitudinal series.

2.3.2.3. WM parcellation

We divided the WM into two distinct parcellation sets: concentric layers and lobar regions, following our previous methodology. ¹² Briefly, for the lobar regions, we used the anatomical Hammers atlas, delineating the frontal, parietal, temporal, occipital, and basal ganglia regions. To define the concentric layers, we computed distance maps from the ventricles to the GM junction (using the GM and CSF maps derived from CAT12v8), thereby segmenting the WM into four layers: Layer 1 is proximal to the ventricles, while Layer 4 encompasses the juxtacortical areas. The resulting parcellations were applied to each time point's FLAIR scan.

2.3.2.4. Extraction of WMH volumes

Total WMH volumes were obtained as the intersection between longitudinal LPA‑derived lesion maps and the WM masks from CAT12v8. Regional WMH volumes were computed from the overlap between the WMH maps and the WM parcellation using an in‐house MATLAB script. To account for differences in MRI acquisition protocols, both total and regional WMH volumes were harmonized using the ComBat harmonization method ²⁴ implemented in the R library neuroCombat.

2.3.2.5. Annualized change calculations

Annualized changes in total and regional Combat‐harmonized WMH volumes were computed as the difference between the last follow‐up and the baseline values, divided by the corresponding time interval between visits, expressed in years. Positive values indicate increases in WMH volume between the two time points, while negative values indicate decreases in WMH volume.

To determine significant changes in WMH volume, we applied a thresholding approach based on a previous study ²⁵ : annual changes greater than 75 mm³/year were considered indicative of progression (i.e., increased WMH volume at follow‐up), while changes inferior to −75 mm³/year were considered indicative of regression (i.e., decreased WMH volume at follow‐up). Changes within the range of −75 mm³/year to 75 mm³/year were considered stable, indicating minimal change in WMH volume.

2.3.2.6. Visual read

To validate the results of the automatic WMH segmentation, an expert neuroradiologist (M.R.A.) performed a visual assessment of paired time points for each subject. Specifically, the rater was blinded to both the subject and the visit order and had to determine whether one scan showed more WMH than the other on a 5‐point Likert scale: −2 (marked decrease), −1 (slight decrease), 0 (no change), +1 (slight increase), +2 (marked increase).

2.3.3. Microbleed segmentation

SWI scans were N4 bias field corrected using ANTs and coregistered to their corresponding T1w image. Microbleed segmentation was manually performed using ITK‐SNAP (v.3.8.0), ²⁶ as previously described. ²⁷

2.4. Blood and cerebrospinal fluid biomarkers

A subset of participants underwent lumbar puncture and/or blood extraction within 1 year of MRI baseline acquisition. ²⁸ , ²⁹ Most participants were screened for apolipoprotein E (APOE) haplotype via Sanger sequencing. We quantified Aβ peptide 1‐40 (Aβ₄₀), Aβ peptide 1‐42 (Aβ₄₂), and tau phosphorylated at threonine 181 (pTau181) in CSF using a commercially available immunoassay in a fully automated platform (Lumipulse; Fujirebio‐Europe), following a previously published protocol. ¹⁶ Additionally, CSF and plasma glial fibrillary acidic protein (GFAP) concentrations were measured using the SR‐X single‐molecule array (Quanterix). CSF neurofilament light chain (NfL) and YKL‐40 (also known as chitinase 3‐like 1) levels were measured with commercially available enzyme‐linked immunosorbent assays (NF‐Light Assay, UmanDiagnostics, and MicroVue EIA, Quidel, respectively) according to the manufacturer's recommendations.

2.5. Statistical analyses

Statistical analyses were conducted using R software (version 4.5.0), and the statistical significance threshold was set at p = 0.05. Baseline demographic differences between DS and HC participants were assessed using the compareGroups library. ³⁰ Chi‐squared tests were used for categorical variables (e.g., sex, APOEε4 status, and vascular risk factors), and Mann–Whitney tests were applied for continuous variables (e.g., age and CSF biomarkers), due to non‐normal distribution.

To examine the association between annual changes of WMH volume and age, we performed locally estimated scatterplot smoothing (LOESS) curves with a tricubic weight function and a span parameter of 0.75, providing an informative representation of the data. The effect of factors sex, APOEε4 status, ID, and AD clinical status on the annual WMH volume changes was assessed using the Mann–Whitney or Kruskal–Wallis tests. For group differences across AD clinical status, Dunn's post‐hoc test was applied with Holm correction for multiple pairwise comparisons, and effect sizes were reported using Rosenthal's r.

Linear robust regression analyses, using the robustbase R library, were used to investigate relationships between annual WMH volume changes and baseline CSF biomarkers of AD, as well as baseline WMH volume. Both baseline WMH volume and fluid biomarker concentrations were log‐transformed to reduce skewness. The impact of microbleed presence on annual WMH changes was evaluated using the Mann‐Whitney test.

3. RESULTS

3.1. Population

In total, we evaluated 449 MRI scans; after a visual quality assessment, 34 of them (7.57%) were excluded due to low transverse resolution or low structural visibility. Then, 91 follow‐up scans (22.64%) were also excluded because of the different acquisition protocols. Lastly, seven scans (2.25%) were excluded due to suboptimal longitudinal coregistration across time points.

Our final sample included 80 adults with DS (n = 178 scans) from the DABNI cohort and 53 HC (n = 124 scans) from the SPIN study. Demographic data at baseline are presented in Table 1. The DS group comprises individuals across the AD continuum, with 65 aDS, 13 sDS (61.55% pDS and 38.45% dDS), and 2 uDS at baseline (see Table S2 for demographic data stratified by clinical diagnosis). Regarding ID level, 26 DS participants had mild ID, 46 moderate ID, and 8 severe/profound ID. The HC group was significantly older than the DS group (median age [interquartile range {IQR}] = 56.91 [12.69] vs. 41.60 [13.9] years, respectively; W = 3826, p < 0.001). There was a trend toward a difference in sex distribution, with more female participants in the HC group (66.04%) compared to the DS group (47.50%, p = 0.054). No significant differences were found for APOEε4 status between groups. Individuals with DS had a lower prevalence of hypertension and dyslipidemia than HC (1.25% vs. 31.8%, p < 0.001; 11.4% vs. 40.9%, p = 0.003), with no difference in diabetes mellitus (2.5% vs. 4.55%, p = 0.521). Individuals with DS demonstrated lower values in CSF for the Aβ42/Aβ40 ratio and higher concentrations of YKL‐40 than HC (p < 0.001; p = 0.003, respectively). No significant differences were found in CSF pTau181, CSF GFAP, and plasma GFAP between groups (p > 0.05).

TABLE 1.

Study participants.

	HC	DS
Parameter	N = 53	N = 80	p‐value
Age at baseline, y	56.91 [51.42;64.11]	41.60 [33.18;47.08]	<0.001
Female	35 (66.04%)	38 (47.50%)	0.054
APOEε4 carriers	18 (34.62%)	17 (26.98%)	0.495
Intellectual disability
Mild		26 (32.50%)
Moderate		46 (57.50%)
Severe/profound		8 (10.00%)
AD clinical diagnosis
aDS		65 (80.00%)
sDS		13 (16.25%)
uDS		2 (3.75%)
Vascular risk factors
Hypertension	7/22 (31.82%)	1/80 (1.25%)	<0.001
Dyslipidemia	9/22 (40.91%)	9/80 (11.39%)	0.003
Diabetes mellitus	1/21 (4.55%)	2/79 (2.50%)	0.521
CAMCOG‐DS		74.00 [61.50;84.00]
MRI data
No. of visits 2/3/4	53 (100.00%) / 15 (28.30%) / 3 (5.66%)	80 (100.00%) / 17 (21.25%) / 1 (1.25%)
Between‐visit time delay, y	2.27 [2.10; 2.60]	2.67 [2.21; 3.54]	0.002
Duration of the follow‐up, y	2.69 [2.20; 4.18]	2.87 [2.26; 4.57]	0.299
WMH measures
WMH at baseline, mm3	2340.79 [1274.43;4387.10]	2701.44 [1371.75;5157.92]	0.261
Annual WMH change, mm3/year	−1.15 [−33.36;68.65]	−10.15 [−90.12;10.69]	0.030
Fluid biomarkers, pg/mL
CSF Aβ42/Aβ40 (N = 33/45)	0.10 [0.10;0.11]	0.08 [0.05;0.09]	<0.001
CSF pTau181 (N = 33/45)	35.80 [26.70;44.00]	36.60 [20.30;69.60]	0.936
CSF NfL(N = 48/51)	382.25 [317.20;551.89]	366.00 [227.10;639.00]	0.190
CSF GFAP (N = 27/50)	3179.94 [1962.01;5246.57]	2869.42 [1669.07;4113.58]	0.411
Plasma GFAP (N = 29/35)	91.47 [72.87;123.45]	92.80 [66.72;156.76]	0.604
CSF YKL40 (N = 53/51)	178.58 [151.84;225.20]	141.41 [79.39;199.86]	0.003

Note: Data are n(%) or median [IQR]. p‐Values indicate the statistical significance of differences between DS and HC participants.

Abbreviations: AD, Alzheimer's disease; aDS, asymptomatic Down syndrome; APOE, apolipoprotein E;Aβ42/Aβ40, concentration ratio between amyloid beta peptide 1‐42 and amyloid beta peptide 1‐40 (pg/mL); CAMCOG‐DS, Cambridge Cognitive Examination for Older Adults with Down Syndrome; CSF, cerebrospinal fluid; DS, Down syndrome; GFAP, glial fibrillary acidic protein concentration (pg/mL); HC, euploid cognitively unimpaired controls; IQR, interquartile range; MRI, magnetic resonance imaging; NfL, neurofilament light chain concentration (pg/mL); pTau181, tau phosphorylated at threonine 181 concentration (pg/mL); sDS, symptomatic Down syndrome; WMH, white matter hyperintensities; YKL‐40, chitinase 3‐like 1.

All participants underwent at least two MRI visits; a subset had a third (17 DS and 15 HC) and a fourth visit (1 DS and 3 HC). The minimum follow‑up duration (baseline to last MRI visit) was 0.56 years for DS and 1.84 years for HC, whereas the maximum follow‑up duration was 9.06 years for DS and 7.03 years for HC. The between‐visit time delay, taking into account all the available intervals, was slightly shorter in HC (2.27 years) than in DS (2.67 years; W = 4464, p = 0.002), while the overall duration of the follow‐up, based on the baseline and last MRI visit, was similar between both groups (HC: 2.69 years; DS: 2.87 years; W = 2346.5, p = 0.299). During follow‐up, 10 aDS participants (16.92%) progressed to sDS. At baseline, there were no significant differences in total WMH volume between the HC and the overall DS groups (HC: median WMH = 2340.79 mm³, DS: 2701.44 mm³; W = 1875.5, p = 0.261). However, when stratifying the DS group into aDS and sDS, we observed significantly higher WMH volumes in sDS compared with both HC and aDS (sDS: 6716 mm³, aDS: 2198.2 mm³, H(2,131) = 16.397, p < 0.001). Annual changes were more pronounced in DS than in HC (‐1.15 mm³/year vs. ‐10.51 mm³/year, W = 2591, p = 0.03).

3.2. Effect of demographic, clinical, and genetic variables on longitudinal WMH

In both DS and HC, WMH exhibited variable changes with age, including both increases and decreases in volume over time (Figure 1A). A higher proportion of individuals with DS presented with decreased WMH compared to HC (26% vs. 9%) and a lower proportion with increased WMH [8% in DS vs. 25% in HC; χ ²(1,133) = 11.08, p = 0.004; Table S3).

FIGURE 1.

Effect of demographic, clinical, and genetic variables on longitudinal WMH. (A) Association between total WMH volume and (B) annual WMH volume with age. The lines/points represent individual participants, with colors indicating their clinical diagnosis. Red points denote participants who exhibited a clinical change between visits along the AD continuum. Shaded areas represent 95% CI: dark gray represents the age‐related change in DS, and light gray in the HC group for visual reference. Boxplots showing the effect of (C) sex, (D) APOE genotype, and (E) intellectual disability on annual WMH volume changes in DS. The light gray background (−75 mm³/year to +75 mm³/year) serves as a visual reference for stable WMH change. AD, Alzheimer's disease; aDS, DS asymptomatic; APOE, apolipoprotein E; DS, Down syndrome; HC, euploid cognitively unimpaired controls; sDS, symptomatic DS; uDS, uncertain DS; and WMH, white matter hyperintensities.

The annual change in WMH was negatively associated with age in the DS group (rho = −0.31, p = 0.006), while this association was not statistically significant in the HC group (rho = 0.07, p = 0.632). The LOESS curve revealed stable annual changes in WMH among individuals with DS until age 40, after which greater variability was observed, and the changes decreased at the group level (Figure 1B).

In individuals with DS, there was no effect of sex (W = 688, p = 0.293), APOEε4 status (W = 438, p = 0.475), or ID [H(2,78) = 2.57, p = 0.276; Figure 1C–E] on the annual change in WMH volume.

3.3. Effect of AD clinical status on the evolution of WMH

The annual changes in total WMH decrease with the progression of AD clinical status in DS [H(2,131) = 11.786, p = 0.003; Figure 2A]. Specifically, sDS exhibited significantly larger decreases in WMH volume compared to aDS (z = 2.71, p = 0.014; aDS: median ΔWMH = −5.39 mm³/y, sDS = −130.83 mm³/y) and HC (z = 3.43, p = 0.002; HC: ΔWMH = −1.15 mm³/y). No significant differences were observed between HC and aDS (z = 1.28, p = 0.201). When examining regional changes, the sDS group showed smaller WMH changes in layer 1 and in the parietal and occipital lobes compared to both aDS and HC, and in layer 2 and the frontal lobe compared to the HC group only (Figure 2B–D; Table S4). Although the sample size was limited, we performed an exploratory analysis to examine annual changes in total WMH by subdividing the sDS group into pDS and dDS subgroups, and by distinguishing within the aDS group those individuals who progressed to sDS at follow‑up (progressors). This analysis revealed greater WMH decreases in the pDS subgroup, which showed significantly larger reductions compared to both HC (z = −4.47, p < 0.001; pDS: ΔWMH = −197.98 mm³/y) and aDS (z = −4.10, p < 0.001; aDS: ΔWMH = 0.33 mm³/y). The progressors (median ΔWMH = −57.95 mm³/y) showed intermediate changes between aDS and pDS, although these differences did not reach statistical significance. In contrast, the dDS subgroup exhibited a significant increase in WMH relative to pDS (z = 2.91, p = 0.03; median ΔWMH = 22.80 mm³/y; Figure S1).

FIGURE 2.

Between‐group differences for total and regional annual WMH volume changes. (A–D) Boxplots illustrate WMH volumes across the brain regions. Significant results at p < 0.05 using the Holm correction. The light gray background (−75 mm³/year to +75 mm³/year) serves as a visual reference for stable WMH change. ns, no significance; ^* p < 0.05; ^** p < 0.01; ^*** p < 0.001. aDS: DS asymptomatic; DS: Down syndrome; HC: euploid cognitively unimpaired controls; sDS: symptomatic DS; and WMH: white matter hyperintensities.

3.4. Association between longitudinal WMH, AD, and SVD markers

In individuals with DS, annual changes in WMH were not significantly associated with any of the fluid biomarkers (all p > 0.05; Figure 3A–F). However, a trend toward mild negative associations was observed with NfL (R²  = 0.04, p = 0.09) and plasma GFAP (R ² = 0.09, p = 0.06).

FIGURE 3.

Association between longitudinal WMH, AD, and SVD markers. Linear robust regression analyses of annual WMH volume changes against (A–F) fluid biomarkers, and (G) baseline WMH volume. The points represent individual participants, with colors indicating their clinical diagnosis. Shaded areas represent 95% CI. (H) Boxplot showing the effect of microbleeds presence. AD, Alzheimer's disease; aDS, DS asymptomatic; Aβ42/Aβ 40, concentration ratio between amyloid β peptide 1–42 and amyloid β peptide 1–40 (pg/mL); CI, confidence interval; CSF, cerebrospinal fluid; DS, Down syndrome; GFAP, glial fibrillary acidic protein concentration (pg/mL); NfL, neurofilament light chain concentration (pg/mL); pTau181, tau phosphorylated at threonine 181 concentration (pg/mL); sDS, symptomatic DS; SVD, small vessel disease; uDS, uncertain DS; WMH, white matter hyperintensities; and YKL‐40, chitinase 3‐like 1 (pg/mL).

Concerning SVD neuroimaging markers, a significant negative association was found between annual changes in WMH and baseline WMH volume in DS (R ² = 0.18, p < 0.001; Figure 3G). Additionally, the presence of at least one cerebral microbleed was associated with lower annual changes in WMH (W = 412, p = 0.024; Figure 3H).

3.5. Sensitivity analyses

Contrary to our expectations, our results demonstrated considerable variability in the evolution of WMH, with a significant proportion of adults with DS showing decreases in WMH volume over time. As these findings were unexpected, we conducted sensitivity analyses to rule out (i) issues related to automatic quantification, and (ii) the confounding effect of WM atrophy. First, we performed a blinded visual assessment, which showed that WMH remained stable in 65% of individuals with DS and 60% of HC. The visual read further showed that 23% of DS and 23% of HC demonstrated a decrease in WMH at follow‐up, whereas 12% of DS and 17% of HC exhibited an increase in WMH volume (Table S5). Specifically, among the aDS group, 72.3% were stable, 18.5% were regressors, and 9.2% were progressors. In the sDS group, 31% were stable, 38% were regressors, and 31% were progressors (Table S5). Second, to ensure that our findings were not merely reflecting progressive WM atrophy, we repeated the primary analyses using WMH volumes adjusted for total WM at the corresponding MRI visit (i.e., WMH/total WM per visit). This adjustment reduced the overall variability and the trajectory of WMH over time (Figure S2). Hence, the association between age and annual change was no longer statistically significant in the DS group (rho = , p = 0.939). However, several DS cases continued to present increased/decreased WMH, and we still observed larger reductions in total WMH volume in the sDS group than in HC, particularly in layer 1 and the parietal and occipital lobes (Figure S3). Results for other demographic variables were consistent, showing no significant effects of sex, APOEε4 status, or ID.

4. DISCUSSION

In the present study, we investigated the evolution of WMH over time in DS. Unlike our hypotheses, we did not observe increases in WMH with age; instead, we found a decrease that was independent of other demographic variables. This reduction in WMH was more pronounced at the symptomatic stage, and was primarily observed in the frontal, parietal, and occipital lobes and the periventricular region (i.e., Layer 1 and Layer 2). Unexpectedly, annual changes in WMH showed no association with any fluid biomarker and were inversely related to baseline WMH volume. Participants with DS who had at least one microbleed exhibited a greater WMH reduction. Visual assessment and adjustment for WM volume confirmed the heterogeneity in WMH trajectories over time in DS and the reduction at symptomatic stages. Together, these results highlight the inconsistent trajectory over time of WMH in DS and provide novel insight into the underlying etiologies of these lesions.

Although cross‐sectional evidence has demonstrated an age‐related increase in WMH in DS, ¹⁰ , ¹¹ , ¹² we found a significant negative association between annual changes in WMH and age in DS. Between the ages of 20 and 40, most individuals with DS exhibited minimal changes in WMH. After age 40, an increasing number of cases experienced significant fluctuations in WMH; primarily decreases, although some increases were also observed. This aligns with a recent study that reported similar fluctuations, both increases and decreases, at older ages in DS, although the association did not reach statistical significance. ³¹ In contrast, in the HC group, the age‐related trajectory showed that WMH burden remained relatively stable until approximately age 55. Beyond this age, most participants continued to show stability, with a few cases demonstrating a substantial increase in WMH volume and a few exhibiting notable decreases. Advancing age has been widely associated with larger WMH volume changes, whereas younger age showed smaller changes over time. ³² , ³³ , ³⁴ , ³⁵ While WMH generally increase with age in euploid individuals, several studies have also reported WMH shrinkage over time. ³² , ³⁶ Population‐based studies have found that approximately 11%–34% of the participants show regression, ³⁷ , ³⁸ , ³⁹ consistent with our results, 9.4% for HC and 26% for DS. We observed no effect of sex, APOEε4 status, or ID on annual WMH changes in DS. This lack of effect is consistent with prior cross‑sectional ¹¹ , ¹² and longitudinal WMH findings, ³¹ as well as evidence from other vascular lesions, ¹¹ , ²⁷ , ⁴⁰ suggesting that sex, APOEε4, and ID have limited impact on SVD features in DS.

We observed significantly greater annual reductions in WMH volumes in DS participants compared to HC, especially in DS individuals at the symptomatic stage of AD (i.e., sDS subgroup). Regional analyses revealed these differences were most prominent in frontal, parietal, and occipital lobes and in periventricular WM regions. A previous longitudinal study in DS reported a monotonic increase in WMH change across diagnostic groups, with negative changes in the asymptomatic and prodromal AD stages and positive rates in more advanced stages. ³¹ The discrepancy with our results might be explained by the fact that we merged the prodromal and dementia stages; indeed, subgroup analyses indicated a positive rate in dementia cases. It is also possible that differences in lesion‐segmentation methods (longitudinal LPA in the present study vs. an in‐house Gaussian mixture model‐based script in Lao et al., 2025) may influence the results. Here, we used a preprocessing pipeline tailored to longitudinal data and additionally performed a blinded visual read. In sporadic AD cohorts, WMH volumes generally increase over time in both mild cognitive impairment (MCI) and dementia, and rates do not significantly differ among groups. ⁴¹ , ⁴² , ⁴³ Moreover, MCI individuals with amyloid pathology accumulate WMH faster than those who are Aβ negative. ⁴⁴ Interestingly, one study reported both greater reductions and increases in WMH in AD patients with a high SVD burden compared to cognitively normal controls and AD with low SVD, ⁴⁵ highlighting potential vascular contribution to the WMH trajectories.

Associations between annual changes in WMH and fluid biomarkers did not reveal robust relationships. This lack of association may reflect heterogeneity across the adult DS lifespan, where non‐linear and stage‑specific biomarker trajectories can mask significant effects. In addition, fluid biomarker data were available only at baseline, preventing change‐change analysis. Interestingly, a greater baseline WMH burden and the presence of cerebral microbleeds, a well‐established neuroradiological feature of CAA, were both associated with a decrease in WMH volume over time. A similar negative relationship between baseline WMH burden and WMH changes was reported by Lao and colleagues. ³¹ These results indicate that the WMH decrease likely reflects dynamics of pathological processes that are more likely in individuals with pre‐existing SVD pathology.

The etiology and pathogenesis of WMH are multifactorial and remain incompletely understood. WMH may, in some instances, reflect irreversible WM damage due to demyelination and axonal loss. ⁴⁶ In DS, we previously reported cross‐sectional associations between WMH and both NfL and GM volume loss, ¹² underscoring a contribution of neurodegeneration to WMH in DS. Here, sensitivity analyses accounting for WM atrophy explained some of the observed WMH reductions, indicating that neurodegeneration partly explains the apparent WMH shrinkage. However, some reductions persisted after adjusting for atrophy, suggesting the involvement of additional mechanisms such as neuroinflammation. There is growing evidence linking inflammation to WMH in DS. ¹² , ⁴⁷ , ⁴⁸ In DS, neuroinflammation may arise from at least three non‐exclusive sources: (i) trisomy 21–related immune dysregulation that promotes chronic neuroinflammation ⁴⁹ ; (ii) AD pathology, as deposition and accumulation of pathological aggregates can exacerbate the inflammatory response ⁵⁰ ; and (iii) CAA, vascular Aβ deposition compromises small‐vessel endothelium, can damage the brain‐blood barrier, and can trigger neuroinflammation. ⁵¹ In some cases, brain‐blood barrier disruption increases permeability, allowing extravasation of interstitial fluid with vasogenic edema; a mechanism consistent with CAA‐related inflammation (CAA‐ri). ⁵² Notably, CAA‐ri has been confirmed in some cases in our DS cohort, including one previously described in a case report. ⁵³ Regardless of cause, neuroinflammation may contribute to the emergence of WMH, but can also partially resolve, leading to subsequent WMH reduction.

As people with DS increasingly participate in clinical trials of anti‐amyloid therapies that can cause cerebral edema and microhemorrhages, rigorously defining the natural history and dynamics of cerebrovascular lesions, such as WMH, becomes essential to guide care. ¹⁴ Our data reveal that WMH changes over time in DS are a non‐linear process. This finding underscores the importance of accounting for this intrinsic variability, not only to avoid confounding interpretations of drug effects but also to exercise caution when WMH are considered as primary or secondary endpoints in clinical trials. Furthermore, a thorough examination of potential pathologic, imaging, demographic, and clinical factors associated with WMH regression could unveil new important targets for intervention.

This study had some limitations. First, the relatively small and unbalanced sample size (n = 65 aDS, 13 sDS, and 2 uDS) reduced statistical power and limited the likelihood of detecting differences in WMH between clinical groups. The limited number of symptomatic participants reflects real‑world constraints on recruiting and scanning adults with DS at advanced AD stages, including reduced MRI tolerability that limits participation and return visits. As a result, a survivor bias is possible as individuals with greater WMH burden are more likely to be symptomatic. Second, while we implemented a specific longitudinal WMH segmentation pipeline and a visual assessment by a neuroradiologist, variability in MRI acquisition parameters may have introduced technical confounding. To minimize intra‑subject variability, we retained only those participants whose MRI visits were all acquired on the same scanner using the same protocol. In addition, we applied a harmonization step to further reduce inter‑subject variability in acquisition parameters across scanners. Third, because most participants had only two MRI timepoints, we were unable to estimate nonlinear rates of change. Future studies using mixed‑effects models, together with additional follow‑up visits over a longer observation window, could capture individual‑level WMH trajectories and potential nonlinearities over time. Finally, the threshold used to classify individuals as exhibiting WMH regression, stability, or progression was adapted from research conducted in non‑DS populations. Such thresholds can influence participant categorization and, consequently, the interpretation of longitudinal WMH changes. ⁵⁴ Future research in DS should aim to establish population‑appropriate thresholds to improve the accuracy, reproducibility, and clinical relevance of WMH quantification in individuals with Down syndrome.

In summary, our findings highlight the variability in the longitudinal evolution of WMH in DS. In younger adults with DS, WMH volumes are low, and annual changes remain minimal, typically close to zero. As individuals age and reach approximately 40–45 years, they exhibit a higher WMH burden, as evidenced by prior cross‐sectional observations, along with greater longitudinal variability, including both increases and decreases over time. Between‐visit decreased WMH was observed more frequently than increased WMH and was related to older age, AD symptoms, and higher baseline WMH. This reduction was only partially explained by the WM atrophy and may suggest mechanisms such as inflammatory resolution. This challenges the assumption that WMH in DS are static accumulations of irreversible injury and underscores the need to better characterize these lesions.

AUTHOR CONTRIBUTIONS

Conceived and designed the study: Alejandra O. Morcillo‐Nieto, Mateus Rozalem‐Aranha, and Alexandre Bejanin. Acquired and interpreted the data: José Enrique Arriola‐Infante, Maria Franquesa‐Mullerat, Sara E. Zsadanyi, Lídia Vaqué‐Alcázar, José Allende Parra, Zili Zhao, Javier Arranz, Isabel Barroeta, Lucia Maure‐Blesa, Laura Videla, Íñigo Rodríguez‐Baz, Laura Del Hoyo Soriano, Bessy Benejam, Susana Fernández, Aida Sanjuan Hernandez, Sandra Giménez, Lucia Pertierra, Daniel Alcolea, Olivia Belbin, Alberto Lleó, María Carmona‐Iragui, and Juan Fortea. Performed the statistical analysis: Alejandra O. Morcillo‐Nieto. Drafted the manuscript, which all authors critically reviewed for important intellectual content: Alejandra O. Morcillo‐Nieto, Mateus Rozalem‐Aranha, and Alexandre Bejanin.

CONFLICT OF INTEREST STATEMENT

JF reported serving on the advisory boards, adjudication committees, or speaker honoraria from AC Immune, Adamed, Alzheon, Biogen, Eisai, Esteve, Fujirebio, Ionis, Laboratorios Carnot, Life Molecular Imaging, Lilly, Novo Nordisk, Perha, Roche, Zambón. DA participated in advisory boards from Fujirebio‐Europe, Roche Diagnostics, Grifols S.A. and Lilly, and received speaker honoraria from Fujirebio‐Europe, Roche Diagnostics, Nutricia, Krka Farmacéutica S.L., Zambon S.A.U., Neuraxpharm, Alter Medica, Lilly and Esteve Pharmaceuticals S.A. JF, DA and OB report holding a patent for markers of synaptopathy in neurodegenerative disease (licensed to ADx NeuroSciences N.V., WO2019175379). MRA is a co‐founder and partner of Masima Soluções em Imagens Médicas LTDA (Porto Alegre, Brazil) and serves an independent consultant to Ionis Pharmaceuticals. MCI has received personal fees for service on the advisory boards, speaker honoraria or educational activities from IMSERSO, Esteve, Lilly, Neuraxpharm, Adium Pharma, and Roche. AL reported receiving personal fees for service on advisory boards, or speaker honoraria from Almirall, Beckman‐Coulter, Biogen, Eisai, Esteve, Fujirebio‐Europe, Grifols, KRKA, Lilly, Novartis, NovoNordisk, Nutricia, Otsuka Pharmaceutical, Roche, and Zambon. AL is co‐author of a patent for markers of synaptopathy in neurodegenerative disease (licensed to ADx, EPI8382175.0) and on antibodies for amyloid precursor, methods and uses thereof European priority (N°EP25382226). SG has received personal fees for service on the advisory boards, speaker honoraria or educational activities from Esteve, Idorsia, Novo Nordisk and Bioproject. JA reported receiving personal fees for service on speaker honoraria or educational activities from Lilly, Esteve, Fujirebio‐Europe and Roche diagnostics. AOMN, LMB, IRB, JEAI, MFM, SEZ, LVA, JAP, ZZ, LV, IB, LDHS, BB, SF, ASH, LP and AB have nothing to disclose. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

All participants and/or their legally authorized representatives gave written informed consent.

Supporting information

Supporting Information

Supporting Information

Morcillo‐Nieto AO, Rozalem‐Aranha M, Maure‐Blesa L, et al. Temporal dynamics of white matter hyperintensities related to Alzheimer's disease in adults with Down syndrome. Alzheimer's Dement. 2026;22:e71157. 10.1002/alz.71157

DATA AVAILABILITY STATEMENT

The authors may share de‐identified data that underlie the results reported in this article. Data will be available upon receipt of a request detailing the study hypothesis and statistical analysis plan. All requests should be sent to the corresponding authors. The steering committee of this study will discuss all requests and decide, based on the novelty and scientific rigor of the proposal, whether data sharing is appropriate. All applicants will be asked to sign a data access agreement.