Diagnostic association of ante-mortem biomarkers with cerebral amyloid angiopathy in a cohort with high Alzheimer's disease comorbidity

Abstract

Background

Cerebral amyloid angiopathy (CAA) frequently co-occurs with Alzheimer's disease (AD), complicating the interpretation of biomarkers. Current diagnostic approaches rely largely on neuroimaging but remain constrained by limited accuracy.

Objective

This study aimed to evaluate the diagnostic discrimination and pathological association of neuropathological scoring systems and fluid biomarkers for CAA using pathologically confirmed samples.

Methods

We examined 100 participants from the Alzheimer's Disease Neuroimaging Initiative with complete neuropathological assessments. Associations and diagnostic performance of amyloid-β (Aβ)-related measures were evaluated using multivariable regression and receiver operating characteristic analyses.

Results

Both Thal phase and the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) score were associated with CAA status. Cerebrospinal fluid (CSF) Aβ₄₂, Aβ₄₀, and Aβ₃₈ concentrations were significantly reduced in CAA, with CSF Aβ₄₂ showing the greatest decrease and highest diagnostic accuracy [area under the curve (AUC) 0.962; sensitivity 88.6%; specificity 100%). Plasma Aβ_42/40 ratio was also markedly lower in CAA and demonstrated robust diagnostic discrimination (AUC 0.913; sensitivity 84.6%; specificity 100%). In contrast, inflammatory and neuronal injury markers did not differ significantly between groups.

Conclusions

Neuropathological scores and fluid biomarkers are associated with the presence and severity of CAA, but this association is confounded by co-existing advanced AD pathology. Despite high discriminative accuracy, the small control sample warrants cautious interpretation of specificity estimates.

Keywords

Alzheimer's disease biomarkers cerebral amyloid angiopathy cerebrospinal fluid neuropathology plasma amyloid-β

Introduction

Cerebral amyloid angiopathy (CAA) is a cerebrovascular disorder marked by amyloid-β (Aβ) deposition in small- to medium-sized cortical and leptomeningeal vessels.^1,2 Its pathogenesis involves impaired Aβ clearance and disruption of vascular integrity, giving rise to clinical manifestations including lobar hemorrhage, cognitive decline, and transient neurological symptoms.^3,4 Characteristic neuroimaging features include cerebral microbleeds, cortical superficial siderosis, and white matter hyperintensities. Although magnetic resonance imaging (MRI) provides characteristic markers of CAA, relying on MRI findings alone offers limited diagnostic accuracy, particularly in patients without hemorrhagic presentation.^5,6

In recent years, cerebrospinal fluid (CSF) and plasma biomarkers have emerged as powerful tools for diagnosing and monitoring neurodegenerative diseases, particularly in tracking dynamic alterations in Aβ metabolism.^7,8 Yet, most studies of CAA have relied on clinical-imaging criteria, such as the modified Boston criteria, rather than neuropathological confirmation, thereby introducing heterogeneity and misclassification bias.^6,9 Although MRI-based standards are widely used in clinical practice, they often diverge from neuropathological benchmarks, the current gold standard, limiting both the accuracy and generalizability of biomarker research.

Using autopsy-confirmed cases from the Alzheimer's Disease Neuroimaging Initiative (ADNI),¹⁰ we systematically evaluated the diagnostic discrimination and pathological associations of neuropathological indices, including Thal phase and Consortium to Establish a Registry for Alzheimer's Disease (CERAD) scores, together with ante-mortem CSF and plasma Aβ biomarkers. By establishing a pathologically verified CAA cohort, we aimed to define biomarker patterns that reliably distinguish CAA from non-CAA cases and reflect disease severity, thereby supporting their application in diagnostic frameworks and clinicopathological research.

Methods

Participants

All data were sourced from the ADNI, a public-private partnership launched in 2003 to assess the diagnostic potential of clinical, imaging, and biochemical biomarkers for AD.¹¹ Data collection and sharing have been approved by all participating ADNI institutions. Ethical approval for ADNI 1, GO, 2, and 3 studies is available on ClinicalTrials.gov under the following identifiers: NCT00106899, NCT01078636, NCT01231971, and NCT02854033, respectively.

This study included 100 participants from the ADNI cohort, all of whom donated their brains for neuropathological analysis at the ADNI Neuropathology Core Laboratory at the Knight Alzheimer's Disease Research Center, Washington University School of Medicine, St Louis. Participants underwent comprehensive clinical cognitive assessments, along with CSF and plasma biomarker measurements. Participants were included solely based on the availability of complete neuropathological assessment data; no additional exclusion criteria were applied. CAA diagnosis was based on the presence of amyloid deposits in cerebral vasculature, confirmed through postmortem brain tissue pathology. Based on these pathological findings, participants were classified into non-CAA and CAA groups.

Measurement of CSF and plasma biomarkers

CSF concentrations of Aβ₄₂, Aβ₄₀, and Aβ₃₈ were measured using the xMAP Luminex platform (Luminex Corp, Austin, TX, USA) and the INNOBIA AlzBio3 kit (Fujirebio, Ghent, Belgium). A panel of 15 inflammatory markers was quantified in CSF using multiplex immunoassays (Millipore Sigma, Burlington, MA, USA). This included tumor necrosis factor receptor 1 (TNFR1), TNFR2, TNFα, transforming growth factor beta (TGFβ1), TGFβ2, TGFβ3, interleukin-6 (IL-6), IL-7, IL-9, IL-10, IL-21, interferon gamma-induced protein 10 (IP-10), IL-12 p40, intercellular adhesion molecule 1 (ICAM-1), and vascular cellular adhesion molecule-1 (VCAM-1), were measured. Plasma neurofilament light chain (NFL) was quantified using the Single Molecule Array (Simoa) method.¹² Plasma Aβ₄₀ and Aβ₄₂ concentrations were quantified using immunoprecipitation followed by enzymatic digestion and liquid chromatography-tandem mass spectrometry (LC-MS/MS) of thawed plasma samples, as previously described.¹³ The plasma Aβ_42/40 ratio was calculated as the concentration of Aβ₄₂ divided by that of Aβ₄₀. Additional details can be found at www.adni-info.org.

Neuropathological examination

Neuropathological assessment followed the NIA-AA guidelines and established procedures,¹⁴ applying three standardized staging systems to define the core features of AD pathology. The Thal phase (0–5) was used to assess the regional distribution of Aβ plaques, the Braak stage (0–6) to map the progression of tau pathology, and the CERAD score (0–3) to grade neuritic plaque density. The overall level of Alzheimer's disease neuropathologic change (ADNC) was summarized using the NIA-AA ABC scoring system, where A reflects Aβ plaque burden (Thal phase), B represents neurofibrillary tangle distribution (Braak stage), and C denotes neuritic plaque density (CERAD score). The composite ABC score classifies ADNC severity as none (0), low (1), intermediate (2), or high (3).

Additional pathological features, such as Lewy bodies, locus ceruleus hypopigmentation, and others, are also assessed at ADNI centers. In accordance with established neuropathological criteria, CAA was graded on a four-tiered semiquantitative scale: None, Mild, Moderate, or Severe.^14,15 This study cohort encompassed this full pathological spectrum. For the analyses herein, we distinguish between CAA status (the binary presence or absence of pathologically confirmed CAA) and CAA stage (the four-tiered severity grading). For detailed operational procedures regarding neuropathological scoring, refer to the National Alzheimer's Coordinating Center's Coding Guidebook for the Neuropathology Form.¹⁴

Neuropsychological assessment

In this study, baseline cognitive function was assessed using the 13-item cognitive subscale of the Alzheimer's Disease Assessment Scale (ADAS-Cog 13) and the Mini-Mental State Examination (MMSE). The ADAS-Cog 13 evaluates various cognitive domains, including memory, attention, language, orientation, and executive function. The total score of ADAS-Cog 13 ranges from 0 to 85, with higher scores indicating more severe cognitive impairment.

Statistical analysis

All statistical analyses were conducted in SPSS (v29.0; IBM) with two-sided significance set at p < 0.05, unless specified otherwise. Data normality was assessed using the Kolmogorov-Smirnov test. Normally distributed variables are reported as mean ± s.d., non-normally distributed variables as median and interquartile range (IQR), and categorical variables as counts and percentages. For two-group comparisons, either t-test (for normally distributed data) or Mann-Whitney U test (for non-normally distributed data) was used. For multiple-group comparisons, one-way ANOVA (for normally distributed data) or Kruskal-Wallis test (for non-normally distributed data) was employed. When multiple comparisons were performed across biomarkers or pathological indices, p values were adjusted using the Benjamini-Hochberg false discovery rate correction to control for false positives. Associations between biomarkers and CAA were assessed using Spearman correlation. To evaluate whether neuropathological scores (Thal, Braak, CERAD) independently predicted CAA, multivariable logistic regression was performed with forward stepwise selection (entry threshold p ≤ 0.05; removal threshold p > 0.1). Discriminative performance of biomarkers for autopsy-confirmed CAA was quantified using receiver-operating characteristic (ROC) analysis, with area under the curve (AUC) and 95% confidence intervals reported.

Results

Sample characteristics

Demographic, clinical and neuropathological features are summarized in Table 1. The cohort comprised 100 participants with neuropathological confirmation, including 17 controls and 83 individuals with CAA. Groups did not differ in age, sex, education or marital status (all p > 0.05). By contrast, APOE ε4 carriage (67.5% versus 23.5%; p = 0.001) and comorbid AD (83.1% versus 47.1%; p = 0.003) were significantly more common in the CAA group.

Table 1.

Clinical and neuropathological characteristics of pathologically-verified CAA.

Characteristics	Number of cases (controls/CAA)	Controls	CAA	p
Age at death, years	17/83	84.9 ± 8.2	82.1 ± 6.8	0.143
Female (n, %)	17/83	5 (29.4%)	19 (22.9%)	0.547
Education, years	17/83	17.0 (13.5–20.0)	16.0 (14.0–18.0)	0.450
Marital status (n, %)	17/83	13 (76.5%)	70 (84.3%)	0.481
APOE ε4 carriers (n, %)	17/83	4 (23.5%)	56 (67.5%)	0.001
AD (n, %)	17/83	8 (47.1%)	69 (83.1%)	0.003
Postmortem interval, hour	17/81	13.0 (5.2–23.6)	8.4 (4.0–17.5)	0.185
Whole brain weight, g	15/79	1264.0 ± 130.6	1202.8 ± 132.5	0.103
Thal phase	17/83	3.0 (0.5–3.5)	5.0 (4.0–5.0)	<0.001
Braak stage	17/83	1.0 (1.0–2.5)	3.0 (3.0–3.0)	<0.001
CERAD score	17/83	0 (0–0)	3.0 (2.0–3.0)	<0.001
ADNC stage	17/83	1.0 (0.5–2.0)	3.0 (3.0–3.0)	<0.001
Moderate/severe cortex atrophy (n, %)	14/69	4 (23.5%)	31 (37.3%)	0.276
Moderate/severe hippocampus atrophy (n, %)	14/71	3 (21.4%)	32 (45.1%)	0.100
Moderate/severe atherosclerosis of the circle of Willis (n, %)	13/67	4 (33.3%)	16 (23.9%)	0.487
Moderate/severe substantia nigra hypopigmentation (n, %)	14/70	0 (0%)	3 (4.3%)	0.100
Moderate/severe locus ceruleus hypopigmentation (n, %)	13/65	0 (0%)	6 (9.2%)	0.582
Moderate/severe neuron loss in the substantia nigra (n, %)	16/81	3 (18.8%)	27 (33.3%)	0.376
Lewy body (n, %)	17/83	9 (52.9%)	43 (51.8%)	0.932
Old infarcts (n, %)	17/82	0 (0%)	8 (9.8%)	0.344
Old microinfarcts (n, %)	17/83	3 (17.6%)	18 (21.7%)	1.000
Old hemorrhages (n, %)	17/81	1 (5.9%)	3 (3.7%)	0.539
Old microhemorrhages (n, %)	17/83	0 (0%)	0 (0%)	1.000
White matter rarefaction (n, %)	17/83	3 (17.6%)	7 (8.4%)	0.367
MMSE	17/82	25.0 (19.5–28.5)	20.0 (15.0–24.0)	0.019
ADAS-Cog 13	16/81	17.1 (11.5–39.5)	34.0 (27.0–46.6)	0.020

CAA: cerebral amyloid angiopathy; APOE: apolipoprotein E; AD: Alzheimer's disease; CERAD: Consortium to Establish a Registry for Alzheimer's Disease; ADNC: Alzheimer's disease neuropathologic change; MMSE: Mini-Mental State Examination; ADAS-Cog 13: 13-item cognitive subscale of the Alzheimer's Disease Assessment Scale.

Neuropathological assessment revealed more severe pathology in CAA, with higher median Thal phase (5.0 versus 3.0), Braak stage (3.0 versus 1.0), CERAD score (3.0 versus 0), and ADNC stage (3.0 versus 1.0) (all p < 0.001). No group differences were observed in postmortem interval, brain weight, cortical or hippocampal atrophy, atherosclerosis, pigment loss in the substantia nigra or locus coeruleus, neuronal loss in the substantia nigra, Lewy body pathology, old infarcts, microinfarcts, hemorrhages or white matter rarefaction (all p > 0.05). Cognitively, individuals with CAA performed worse, showing lower MMSE scores (median 20.0 versus 25.0, p = 0.019) and higher ADAS-13 scores (median 34.0 versus 17.1, p = 0.020).

Independent associations with pathologically-verified CAA

In multivariable logistic regression analyses, both Thal phase and CERAD score remained significantly associated with the presence of CAA after adjustment for covariates (Table 2). Thal phase remained significant in both models (adjusted OR = 2.541, 95% CI: 1.058–6.103; p = 0.036) and Model 2 (adjusted OR = 3.312, 95% CI: 1.257–8.725; p = 0.015). CERAD score also showed strong associations (adjusted OR = 9.128, 95% CI: 1.858–44.825; p = 0.006) and Model 2 (adjusted OR = 8.140, 95% CI: 1.608–41.205; p = 0.011). By contrast, Braak stage, ADNC stage, comorbid AD, and APOE ε4 status were not associated with CAA (all p > 0.05).

Table 2.

Multivariable effect of the pathological and cognitive factors on pathologically-verified CAA.

Variable	Model 1		Model 2
Variable	Adjusted OR (95% CI)	p	Adjusted OR (95% CI)	p
Thal phase	2.541 (1.058–6.103)	0.036	3.312 (1.257–8.725)	0.015
Braak stage	0.726 (0.129–4.069)	0.715	0.668 (0.077–5.179)	0.62
CERAD score	9.128 (1.858–44.825)	0.006	8.140 (1.608–41.205)	0.011
ADNC stage	0.355 (0.037–3.352)	0.366	0.274 (0.021–3.572)	0.323
AD	6.740 (0.756–60.034)	0.087	10.428 (0.990–109.734)	0.051
APOE ε4	1.640 (0.278–9.649)	0.584	2.468 (0.346–17.596)	0.367
MMSE	1.325 (1.056–1.663)	0.015	-	-
ADAS-Cog 13	-	-	0.890 (0.815–0.973)	0.010

Among cognitive measures, MMSE was associated with CAA in Model 1 (adjusted OR = 1.325, 95% CI: 1.056–1.663; p = 0.015), whereas ADAS-13 contributed independently in Model 2 (adjusted OR = 0.890, 95% CI: 0.815–0.973; p = 0.010). Consistent with these results, Thal phases were elevated across mild, moderate and severe CAA compared with controls (Figure 1A), and CERAD scores were higher in both mild and severe CAA groups (Figure 1B; all p < 0.05).

Figure 1.

Distribution of Thal phase and CERAD score across CAA severity stages. Distribution of Thal phase (A) and CERAD score (B) across different CAA severity levels. Significant differences between the groups were assessed using Kruskal-Wallis test and statistical significance indicated by asterisks after false discovery rate correction (*p < 0.05, **p < 0.01). Small violin plots represent the distribution of scores, with the width of each plot indicating the sample density at various points. The dashed lines mark the 25th percentile, median, and 75th percentile of the data. CERAD: Consortium to Establish a Registry for Alzheimer's Disease; CAA: cerebral amyloid angiopathy.

Biomarker characteristics and diagnostic accuracy in CAA

Building on the neuropathological findings (Table 2), which established Aβ pathology as an independent correlate of CAA, we next examined whether ante-mortem CSF biomarkers could discriminate CAA from controls (Table 3). CSF concentrations of Aβ₄₂, Aβ₄₀, and Aβ₃₈ were all reduced in the CAA group (all p < 0.01), with Aβ₄₂ showing the largest decrease (766.1 ± 507.3 pg/mL versus 2262.5 ± 966.1 pg/mL in controls; p < 0.001). Consistently, the plasma Aβ_42/40 ratio was lower in CAA (0.10 ± 0.01 versus 0.132 ± 0.004; p = 0.020). By contrast, CSF levels of inflammatory mediators and adhesion molecules (TNFR1, TNFR2, TGFβ1-3, IL-6, IL-7, IL-9, IL-10, IL-21, TNFα, IP-10, IL-12p40, ICAM-1, VCAM-1) did not differ between groups (all p > 0.05), nor did plasma NFL concentrations (p = 0.637).

Table 3.

Differences in ante mortem biomarkers between pathologically-verified CAA and controls group.

Characteristics	Number of cases (controls/CAA)	Controls	CAA	p
CSF Aβ₄₂ (pg/ml)	6/35	2262.5 ± 966.1	766.1 ± 507.3	<0.001
CSF Aβ₄₀ (pg/ml)	6/35	11598.8 ± 3972.9	7691.0 ± 2439.4	0.002
CSF Aβ₃₈ (pg/ml)	6/35	2613.0 ± 867.0	1792.6 ± 573.1	0.004
CSF TNFR1 (pg/ml)	9/28	856.8 (721.4–1205.4)	806.7 (709.7–963.0)	0.272
CSF TNFR2 (pg/ml)	9/28	1052.0 (826.2–1222.5)	1017.5 (855.1–1281.6)	0.831
CSF TGFβ1 (pg/ml)	9/28	105.2 (87.3–144.1)	101.9 (87.3–124.3)	0.595
CSF TGFβ2 (pg/ml)	9/28	164.4 (136.5–180.4)	169.1 (141.3–210.1)	0.523
CSF TGFβ3 (pg/ml)	9/28	3.1 (2.6–3.5)	2.8 (2.5–3.7)	0.501
CSF IL-6 (pg/ml)	9/28	4.5 (3.2–5.1)	4.2 (2.9–6.1)	0.957
CSF IL-7 (pg/ml)	9/28	0.9 (0.5–1.9)	0.9 (0.4–1.6)	0.762
CSF IL-9 (pg/ml)	9/28	3.2 (2.3–5.0)	3.3 (2.5–4.3)	0.974
CSF IL-10 (pg/ml)	9/28	5.8 (3.4–7.3)	5.2 (3.9–7.6)	0.763
CSF IL-21 (pg/ml)	9/28	8.0 (3.1–23.7)	3.6 (0.4–11.9)	0.062
CSF TNFα (pg/ml)	9/28	1.9 (1.4–2.6)	1.8 (1.4–2.1)	0.670
CSF IP-10 (pg/ml)	9/28	5102.5 (4595.2–9414.1)	4508.6 (3064.3–6325.7)	0.172
CSF IL-12 p40 (pg/ml)	9/28	1.1 (0.1–1.8)	1.3 (0.4–2.4)	0.545
CSF ICAM-1 (pg/ml)	9/28	349.3 (170.0–477.1)	348.2 (265.5–681.1)	0.304
CSF VCAM-1 (pg/ml)	9/28	46108.3 (25176.2–68252.0)	35405.3 (27733.1–56235.5)	0.723
Plasma Aβ_42/40 (%)	4/13	0.132 ± 0.004	0.1 ± 0.01	0.020
Plasma NFL (pg/ml)	14/51	58.4 (43.1–76.6)	54.7 (45.7–71.3)	0.637

CAA: cerebral amyloid angiopathy; CSF: cerebrospinal fluid; Aβ: amyloid-β; TNFR: tumor necrosis factor receptor; TGFβ: transforming growth factor beta; IL: interleukin; TNFα: tumor necrosis factor alpha; ICAM-1: intercellular adhesion molecule 1; VCAM-1: vascular cellular adhesion molecule-1; NFL: neurofilament light chain.

Spearman correlation analysis was conducted to assess the associations of CAA status and pathological stage with CSF biomarkers and neuropathological scores, as summarized in the heatmap (Figure 2). CSF Aβ₄₂ showed strong negative correlations with both CAA status (r_s = −0.565) and CAA stage (r_s = −0.675). Similarly, plasma Aβ_42/40 ratio was negatively correlated with CAA (r_s = −0.594) and CAA stage (r_s = −0.666). Moderate negative correlations were observed for CSF Aβ₄₀ and CSF Aβ₃₈ with both CAA and its severity stage. In contrast, the CERAD score was positively correlated with CAA (r_s = 0.344) and CAA stage (r_s = 0.329). The Thal phase also showed a positive correlation with CAA (r_s = 0.375) and CAA stage (r_s = 0.437).

Figure 2.

Spearman correlation heatmap of biomarkers and neuropathological scores with CAA and CAA stages. A Heatmap depicts Spearman correlation coefficients (r_s) for the association between CAA status, CAA pathological stage, and various biomarkers or neuropathological scores. Red hues represent positive correlations; blue hues represent negative correlations. Color intensity and the numerical value within each cell indicate the strength of the correlation. All shown correlations were significant (p < 0.05). CERAD: Consortium to Establish a Registry for Alzheimer's Disease; CAA: cerebral amyloid angiopathy; CSF: cerebrospinal fluid; Aβ: amyloid- β.

As shown in Figure 3, the diagnostic utility of several biomarkers for pathologically confirmed CAA was evaluated. CSF Aβ₄₂ demonstrated exceptional discriminative capacity, with an AUC of 0.962 (95% CI: 0.850–0.997). Using an optimal cut-off of 1020.0 pg/mL, it achieved a sensitivity of 88.57% and specificity of 100.0%. CSF Aβ₄₀ and Aβ₃₈ also exhibited significant diagnostic value, with AUCs of 0.795 (95% CI: 0.640–0.905) and 0.771 (95% CI: 0.614–0.888), respectively. At cut-offs of 10699 pg/mL and 2438.0 pg/mL, CSF Aβ₄₀ provided a sensitivity of 94.29% and specificity of 66.67%, while CSF Aβ₃₈ yielded a sensitivity of 94.28% and specificity of 66.57%. The plasma Aβ_42/40 ratio also showed high discriminatory power, achieving an AUC of 0.913 (95% CI: 0.675–0.994). At a cut-off of 0.125, it attained a sensitivity of 84.62% and specificity of 100.00%. Given the small number of controls, specificity estimates should be interpreted with caution.

Figure 3.

Receiver operating characteristic curves for biomarkers in diagnosing CAA. CSF: cerebrospinal fluid; Aβ: amyloid-β; AUC: area under the curve.

Discussion

Our study demonstrates that both Thal phase and CERAD score were statistically associated with the presence of CAA. In parallel, CSF Aβ concentrations showed strong inverse relationships with disease severity, suggesting their utility as dynamic markers of disease progression. Notably, both CSF and plasma Aβ measures exhibited high discriminative accuracy for CAA within this highly selected, AD-enriched autopsy cohort, supporting their potential as clinically accessible biomarkers for diagnosis and stratification. Both Thal phase and CERAD scores, which primarily index parenchymal amyloid deposition, are also influenced by the presence of AD pathology, which is frequently comorbid with CAA. Thus, the biomarkers identified in this study may not exclusively reflect vascular amyloid deposition in CAA, but may also capture broader amyloid pathology in AD.

The hallmark of CAA is the abnormal deposition of Aβ within the walls of small vessels in the cortex and leptomeninges.¹⁶ The Thal staging system, which tracks the neuroanatomical progression of Aβ deposition, closely mirrors the vascular Aβ accumulation pattern characteristic of CAA.¹⁷ Elevated Thal stages in CAA patients suggest a more extensive cerebrovascular Aβ burden, and higher stages are generally associated with more severe disease progression.¹⁸ Additionally, data from the Cognitive Function and Ageing Study indicate that capillary involvement (CAA type I) is significantly correlated with higher Thal and Braak stages.¹⁹ Further neuropathological analysis of 1113 community-dwelling older adults has confirmed that CAA is an independent risk factor for dementia.²⁰ In contrast, the CERAD score quantifies the local density of neuritic plaques, providing insight into the propensity of Aβ aggregation within the brain parenchyma.²¹ The independent predictive value of both the Thal stage and the CERAD score suggests that the risk of CAA is influenced not only by the presence of Aβ, but also by the interaction between its spatial distribution (captured by the Thal stage) and the density of its deposition (as reflected by the CERAD score).^18,22 Although APOE ε4 was not an independent predictor, its biological relevance is evident. The ε4 allele is associated with increased Aβ deposition and non-hemorrhagic CAA, while ε2 may predispose to severe cortical superficial siderosis.²³

In this study, we observed a significant inverse correlation between CSF Aβ levels and the severity of CAA. Previous reports have shown that CAA patients exhibit a 22.5% reduction in CSF Aβ₄₀ and a 41.8% reduction in Aβ₄₂ compared to healthy controls, with the extent of reduction strongly correlating with CAA pathological grading.²⁴ Furthermore, CAA patients may experience a more pronounced decrease in Aβ₄₀ levels compared to those with AD.²⁵ A meta-analysis has further confirmed that CSF Aβ₄₀ levels are inversely correlated with CAA severity, independent of comorbid AD pathology.²⁶ Our analysis of ante-mortem CSF biomarkers in confirmed CAA cases also revealed that CSF levels of Aβ₃₈, Aβ₄₀, and Aβ₄₂ are significantly associated with both the presence and severity of CAA. Recent CSF proteomic studies suggest that CAA is associated with distinct molecular patterns beyond Aβ_42/40, including reductions in synaptic and granin-related proteins, as well as lower levels of Aβ₃₈, Aβ₄₀, and Aβ₄₂.^27–29 Additionally, we found that the plasma Aβ_42/40 ratio strongly and inversely correlates with CAA status and disease stage. In CAA, Aβ deposition in the walls of cerebral vessels impairs vascular integrity and clearance mechanisms. This reduction in Aβ availability in the interstitial fluid leads to decreased concentrations in the CSF and consequently alters plasma levels through fluid exchange, reflecting the dynamic equilibrium of Aβ along the interstitial fluid-CSF-plasma pathway.^30,31 However, it is crucial to consider that the observed changes may also be influenced by co-existing AD pathology, particularly parenchymal Aβ plaques, which complicates the interpretation of these biomarkers as specific to CAA. Additionally, this impairment may be compounded by coexisting AD pathology, such as parenchymal Aβ plaques. The combined effects of vascular and parenchymal amyloid deposition help explain the robust inverse correlation between Aβ_42/40 levels and CAA severity, highlighting the utility of these biomarkers as dynamic indicators of the brain's overall Aβ burden.

This study demonstrates that CSF and plasma Aβ biomarkers effectively differentiate patients with CAA from healthy controls. A recent meta-analysis, which included 327 CAA patients and 336 healthy controls, observed significantly lower CSF Aβ₄₀ levels in the CAA group compared to controls.³² A previous study of 67 patients with probable CAA also reported that CSF Aβ₄₀ and Aβ₄₂ exhibited high sensitivity and specificity in distinguishing CAA from healthy controls.²⁴ Another investigation involving 63 CAA patients indicated that CSF Aβ₄₂ achieved an accuracy of 89% in distinguishing CAA from controls, significantly outperforming the plasma Aβ_42/40 ratio.²⁵ In contrast to these earlier studies, the current work, based on pathologically confirmed CAA cases, found that the plasma Aβ_42/40 ratio achieved a discrimination accuracy of 91%. This discrepancy may stem from the fact that all CAA cases in our cohort were pathologically verified, whereas previous studies primarily relied on clinical and imaging diagnoses. Recent research also suggests that plasma proteomic biomarkers perform excellently in differentiating CAA from controls, and risk stratification models built around key proteins can accurately identify patients at high risk of incident lobar hemorrhage.³³ Compared to other reports, our pathologically confirmed CAA cohort with high AD comorbidity showed CSF Aβ₄₂ levels similar to mixed CAA-AD/AD groups, and intermediate Aβ₄₀ levels.³⁴ This reflects cohort differences: the cognitive-onset ADNI sample contrasts with cohorts that include purer, hemorrhagic CAA cases. Our findings are therefore complementary, revealing the biomarker signature of CAA is heavily modulated by concurrent AD pathology and clinical presentation.

We observed no significant intergroup differences in inflammatory or neurodegenerative markers. This likely reflects the confinement of CAA-associated inflammation to perivascular regions, consistent with prior reports finding no association between CAA and systemic factors such as TGFβ1 or interleukins.^35,36 Furthermore, while NFL is a putative marker for late-stage CAA,³⁷ its lack of elevation here may stem from study design; specifically, relying on single-time-point baseline plasma from subjects with advanced neuropathology may have obscured dynamic, stage-dependent signals.

Several major limitations, rooted in the nature of our study cohort, must be explicitly acknowledged. A primary limitation of our study is the significant imbalance in comorbid AD pathology between groups, a profound confounding factor that complicates efforts to disentangle CAA-specific biomarker effects from the overall parenchymal amyloid load. Although we performed multivariate regression, the high degree of collinearity among amyloid-related predictors limits the reliable inference of independent associations. More rigorous statistical methods, such as propensity score matching, were precluded by our sample size. Consequently, our conclusions regarding biomarker performance must be interpreted with considerable caution, as they likely reflect a mixed CAA-AD pathological state. Secondly, the limited sample size and pathological heterogeneity of AD restricted our capacity for in-depth subgroup analyses. Furthermore, the cross-sectional, postmortem design entirely precludes any assessment of longitudinal biomarker dynamics, and the influence of unmeasured confounders can never be fully ruled out. It is crucial to emphasize that this design only validates retrospective diagnostic associations and inherently lacks predictive value. Therefore, the analytical framework cannot be extrapolated to predict disease progression or clinical outcomes.

Conclusions

Our findings demonstrate that Thal phase and CERAD score are strongly associated with CAA and that CSF and plasma Aβ biomarkers effectively distinguish CAA from healthy controls. the overlapping amyloid pathology between CAA and AD necessitates cautious interpretation. Plasma Aβ_42/40 demonstrates high diagnostic accuracy and may serve as a risk stratification tool in specific contexts, though its value for community screening requires prospective validation in unselected populations.

Footnotes

Acknowledgements

Data collection and sharing for this project was funded by the ADNI (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer's Association; Alzheimer's Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer's Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California. The data used in this study, which was being prepared, was obtained from the ADNI database. Therefore, researchers within ADNI had participated in the design and implementation of ADNI and/or had provided data, but they had not been involved in the analysis or writing of this report. The complete list of ADNI investigators can be found at the following location:

ORCID iD

Lingling Huang

Ethical considerations

This study used publicly available datasets.

Consent to participate

Not applicable

Consent for publication

Not applicable

Author contribution(s)

Kemi Hou: Data curation; Formal analysis; Methodology; Writing – original draft.

Haojun Lu: Data curation; Formal analysis; Methodology; Writing – original draft.

Rong Zhang: Investigation; Methodology; Validation.

Yadong Wang: Formal analysis; Investigation.

Ge Xu: Formal analysis; Software.

Juan Lu: Data curation; Formal analysis; Methodology; Writing – original draft.

Dong Xing: Data curation; Investigation; Validation.

Lingling Huang: Conceptualization; Investigation; Supervision; Writing – review & editing.

Yaming Sun: Conceptualization; Funding acquisition; Supervision; Writing – review & editing.

Funding

This work was supported by Suzhou major clinical diseases diagnosis and treatment technology special (LCZX202121) and Research Project on Social Development Technological Innovation of Zhangjiagang Science and Technology Bureau (ZKS2138).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

Researchers interested in accessing the ADNI data can do so through the LONI Imaging & Data Archive. To apply for access, they can visit the ADNI website at .

References

Charidimou

Boulouis

Gurol

, et al. Emerging concepts in sporadic cerebral amyloid angiopathy. Brain 2017; 140: 1829–1850.

van Veluw

Benveniste

Bakker

, et al. Is CAA a perivascular brain clearance disease? A discussion of the evidence to date and outlook for future studies. Cell Mol Life Sci 2024; 81: 239.

Koemans

Chhatwal

van Veluw

, et al. Progression of cerebral amyloid angiopathy: a pathophysiological framework. Lancet Neurol 2023; 22: 632–642.

Kozberg

Perosa

Gurol

, et al. A practical approach to the management of cerebral amyloid angiopathy. Int J Stroke 2021; 16: 356–369.

Fandler-Höfler

Obergottsberger

Ambler

, et al. Association of the presence and pattern of MRI markers of cerebral small vessel disease with recurrent intracerebral hemorrhage. Neurology 2023; 101: e794–e804.

Switzer

Charidimou

McCarter

, et al. Boston Criteria v2.0 for cerebral amyloid angiopathy without hemorrhage: an MRI-neuropathologic validation study. Neurology 2024; 102: e209386.

Margraf

Jensen-Kondering

Weiler

, et al. Cerebrospinal fluid biomarkers in cerebral amyloid angiopathy: new data and quantitative meta-analysis. Front Aging Neurosci 2022; 14: 783996.

Hsieh

Tsai

Liu

, et al. Plasma phosphorylated tau 217 as a discriminative biomarker for cerebral amyloid angiopathy. Eur J Neurol 2025; 32: e70066.

Greenberg

Charidimou

. Diagnosis of cerebral amyloid angiopathy: evolution of the Boston criteria. Stroke 2018; 49: 491–497.

10.

Veitch

Weiner

Aisen

, et al. Using the Alzheimer's Disease Neuroimaging Initiative to improve early detection, diagnosis, and treatment of Alzheimer's disease. Alzheimers Dement 2022; 18: 824–857.

11.

Weiner

Kanoria

Miller

, et al. Overview of Alzheimer's Disease Neuroimaging Initiative and future clinical trials. Alzheimers Dement 2025; 21: e14321.

12.

Shen

Wang

, et al. Plasma amyloid, tau, and neurodegeneration biomarker profiles predict Alzheimer's disease pathology and clinical progression in older adults without dementia. Alzheimers Dement (Amst) 2020; 12: e12104.

13.

Ovod

Ramsey

Mawuenyega

, et al. Amyloid β concentrations and stable isotope labeling kinetics of human plasma specific to central nervous system amyloidosis. Alzheimers Dement 2017; 13: 841–849.

14.

Montine

Phelps

Beach

, et al. National Institute on Aging-Alzheimer's Association guidelines for the neuropathologic assessment of Alzheimer's disease: a practical approach. Acta Neuropathol 2012; 123: 1–11.

15.

Olichney

Hansen

Hofstetter

, et al. Association between severe cerebral amyloid angiopathy and cerebrovascular lesions in Alzheimer disease is not a spurious one attributable to apolipoprotein E4. Arch Neurol 2000; 57: 869–874.

16.

Weidauer

Hattingen

. Cerebral amyloid angiopathy: clinical presentation, sequelae and neuroimaging features-an update. Biomedicines 2025; 13: 603.

17.

Murray

Lowe

Graff-Radford

, et al. Clinicopathologic and ¹¹C-Pittsburgh compound B implications of Thal amyloid phase across the Alzheimer's disease spectrum. Brain 2015; 138: 1370–1381.

18.

Thal

Gawor

. Cerebral amyloid angiopathy: neuropathological diagnosis, link to Alzheimer's disease and impact on clinics. Clin Neuropathol 2023; 42: 176–189.

19.

Wharton

Wang

Parikh

, et al. Epidemiological pathology of Aβ deposition in the ageing brain in CFAS: addition of multiple Aβ-derived measures does not improve dementia assessment using logistic regression and machine learning approaches. Acta Neuropathol Commun 2019; 7: 198.

20.

Boyle

Nag

, et al. Cerebral amyloid angiopathy and cognitive outcomes in community-based older persons. Neurology 2015; 85: 1930–1936.

21.

Trejo-Lopez

Yachnis

Prokop

. Neuropathology of Alzheimer's disease. Neurotherapeutics 2022; 19: 173–185.

22.

Wheeler

Irimia

Braskie

. Using neuroimaging to study cerebral amyloid angiopathy and its relationship to Alzheimer's disease. J Alzheimers Dis 2024; 97: 1479–1502.

23.

Theodorou

Tsantzali

Kapaki

, et al. Cerebrospinal fluid biomarkers and apolipoprotein E genotype in cerebral amyloid angiopathy. A narrative review. Cereb Circ Cogn Behav 2021; 2: 100010.

24.

Sembill

Lusse

Linnerbauer

, et al. Cerebrospinal fluid biomarkers for cerebral amyloid angiopathy. Brain Commun 2023; 5: fcad159.

25.

Grangeon

Paquet

Guey

, et al. Cerebrospinal fluid profile of tau, phosphorylated tau, Aβ₄₂, and Aβ₄₀ in probable cerebral amyloid angiopathy. J Alzheimers Dis 2022; 87: 791–802.

26.

Charidimou

Friedrich

Greenberg

, et al. Core cerebrospinal fluid biomarker profile in cerebral amyloid angiopathy: a meta-analysis. Neurology 2018; 90: e754–e762.

27.

Vervuurt

Schrader

de Kort

, et al. Cerebrospinal fluid shotgun proteomics identifies distinct proteomic patterns in cerebral amyloid angiopathy rodent models and human patients. Acta Neuropathol Commun 2024; 12: 6.

28.

De Kort

Kuiperij

Kersten

, et al. Normal cerebrospinal fluid concentrations of PDGFRβ in patients with cerebral amyloid angiopathy and Alzheimer's disease. Alzheimers Dement 2022; 18: 1788–1796.

29.

De Kort

Kuiperij

Marques

, et al. Decreased cerebrospinal fluid amyloid β 38, 40, 42, and 43 levels in sporadic and hereditary cerebral amyloid angiopathy. Ann Neurol 2023; 93: 1173–1186.

30.

Greenberg

Bacskai

Hernandez-Guillamon

, et al. Cerebral amyloid angiopathy and Alzheimer disease - one peptide, two pathways. Nat Rev Neurol 2020; 16: 30–42.

31.

Holth

Fritschi

Wang

, et al. The sleep-wake cycle regulates brain interstitial fluid tau in mice and CSF tau in humans. Science 2019; 363: 880–884.

32.

Theodorou

Tsantzali

Stefanou

, et al. CSF and plasma biomarkers in cerebral amyloid angiopathy: a single-center study and a systematic review/meta-analysis. Eur Stroke J 2025; 10: 278–288.

33.

Sha

, et al. Circulating proteomic biomarkers for cerebral amyloid angiopathy screening and risk stratification. Alzheimers Dement 2025; 21: e70044.

34.

Losa

Ramusino

Cama

, et al. Cerebrospinal fluid biomarkers profiling in cerebral amyloid angiopathy and relationship with disease phenotypes. J Am Heart Assoc 2025; 14: e044784.

35.

Greenberg

Cho

O'Donnell

, et al. Plasma beta-amyloid peptide, transforming growth factor-beta 1, and risk for cerebral amyloid angiopathy. Ann N Y Acad Sci 2000; 903: 144–149.

36.

Kimura

Takemura

Saito

, et al. Comparison of cerebrospinal fluid profiles in Alzheimer's disease with multiple cerebral microbleeds and cerebral amyloid angiopathy-related inflammation. J Neurol 2017; 264: 373–381.

37.

Rasing

Voigt

Koemans

, et al. Serum and cerebrospinal fluid neurofilament light chain and glial fibrillary acid protein levels in early and advanced stages of cerebral amyloid Angiopathy. Alzheimers Res Ther 2024; 16: 86–97.