Alzheimer's disease (AD), the leading cause of dementia, is initiated by excessive accumulation of amyloid β (Aβ) as senile plaques within the brain parenchyma [1].　Although Aβ accumulation is observed even in cognitively normal individuals, it is significantly prominent in patients with AD and typically begins 20 to 30 years prior to dementia onset [2]. Aβ accumulation is not solely confined to senile plaques (plaque amyloid); it also deposits on blood vessels, forming vascular amyloid [3]. This vascular amyloid deposition, referred to as cerebral amyloid angiopathy of the Aβ type (CAA), is observed in up to 48% of patients with AD [4]. Patients with CAA frequently develop hemorrhage and edema in the brain due to vascular disruption or increased permeability, which is partly due to an immune response to the deposited vascular amyloid [5]. Hemorrhagic and edematous imaging abnormalities associated with CAA have been described by terms such as cortical and subcortical microbleeds (MBs), cortical and subcortical superficial siderosis (SS), and CAA-related inflammation (CAAri). These imaging abnormalities have gained attention in dementia research due to their frequent occurrence as complications of anti-amyloid passive immunotherapy (anti-amyloid therapy). Consequently, the term amyloid-related imaging abnormalities (ARIA) was introduced to encompass idiopathic and iatrogenic CAA-related imaging abnormalities [6]. Among the multiple mechanisms underlying ARIA, particularly ARIA-E, which manifests as vasogenic edema, vascular amyloid deposition may increase the susceptibility to vascular injury following anti-amyloid therapy by impairing clearance mechanisms and enhancing immune-mediated inflammation. It is hypothesized that anti-amyloid antibodies targeting vascular Aβ deposits may provoke local inflammatory responses, leading to increased vascular permeability and resultant edema or microhemorrhage [7]. Therefore, noninvasive and accurate assessment of vascular amyloid burden is crucial for predicting and reducing the risk of ARIA during anti-amyloid therapy.

Histopathological confirmation through brain biopsy represents the gold standard for diagnosing CAA; however, this approach is clinically infeasible [8]. Hemorrhagic changes, such as microbleeds (MBs) and superficial siderosis (SS) detected via magnetic resonance imaging (MRI), are reliable imaging biomarkers for identifying CAA under the Boston criteria version 2.0 [8]. Nevertheless, these biomarkers reflect secondary consequences of CAA rather than the presence or the severity of the vascular amyloid itself [9]. Amyloid PET imaging identifies amyloid accumulation in the brain; however, it does not distinguish between vascular amyloid and plaque amyloid [10]. Cerebrospinal fluid (CSF) levels of Aβ-related species decline as amyloid accumulates in the brain, and these decline profiles differ between CAA and AD. In AD, CSF Aβ42 levels are reduced compared to controls, while CSF Aβ40 levels remain largely unchanged [11]. This reduction is often attributed to the adsorption of Aβ42 in the cerebral interstitial fluid onto plaque amyloid, which primarily consists of insoluble, long-chain Aβ molecular species like Aβ42 [12]. Conversely, vascular amyloid contains higher concentrations of shorter Aβ peptides (e.g., Aβ40), and in brains with abundant vascular amyloid, both Aβ42 and Aβ40 are thought to be sequestered into the vascular deposits. As a result, CAA reduces both CSF Aβ42 and Aβ40 levels [9]. Decreases in CSF Aβ40 levels are correlated with increased lobar MB count, greater white matter hyperintensity volume, and the presence of cortical SS [13]. Therefore, in patients with AD or MCI due to AD, the presence of decreased CSF Aβ40 in addition to reduced CSF Aβ42 may suggest a relatively greater burden of vascular amyloid deposition in the brain. However, because collecting CSF is invasive, it is not suitable for regular use as a biomarker or for tracking changes over time during anti-amyloid treatment.

MRI-visible perivascular spaces in the centrum semiovale (CSO-PVS), which have been incorporated as a non-hemorrhagic MRI biomarker in the Boston criteria version 2.0, provide a noninvasive method for assessing vascular amyloid[8]. The diagnostic value of CSO-PVS findings has been demonstrated in patients with CAA-related hemorrhagic abnormalities [14]. However, patients with AD receiving anti-amyloid therapy, who are the primary subjects for predicting ARIA onset, generally exhibit negligible CAA-related hemorrhagic findings, particularly at the commencement of treatment. Whether CSO-PVS findings can be adapted to reflect vascular amyloid accumulation in amyloid pathology-positive individuals with such minimal CAA-associated hemorrhagic manifestations remains an open question. Establishing a method to evaluate the utility of CSO-PVS findings for detecting vascular amyloid deposition in Alzheimer's disease (AD) patients without overt hemorrhagic manifestations associated with CAA is a challenging task. This difficulty arises from ethical constraints, as it is problematic to compare CSO-PVS findings with brain pathology in relatively healthy individuals. Therefore, this study indirectly addresses the issue by analyzing the relation between CSO-PVS grades and reductions in both CSF Aβ42 and Aβ40 in patients evaluated for anti-amyloid therapy eligibility (Fig. 1). Combining CSF Aβ42 and Aβ40 concentrations into a composite index may provide a more sensitive measure of amyloid pathology, especially in cases where each individual marker exhibits borderline significance. Previous studies have suggested that using ratios or multiplication products can enhance the sensitivity and robustness of CSF biomarkers by accounting for inter-individual variability and assay noise [15]. Therefore, we also explored the ratio (Aβ42/40 ratio) and the product of Aβ42 and Aβ40 (Aβ40 × Aβ42) as potential surrogate indicators of plaque amyloid and vascular amyloid burden, respectively. The latter approach may better reflect amyloid burden in the presence of vascular involvement by emphasizing the simultaneous reductions of both Aβ species.

All participants underwent comprehensive evaluations, including medical history interviews and general and neurological examinations conducted by trained neurologists (TK, FKM) to exclude comorbidities such as parkinsonism, including DLB and iNPH. CDR and MMSE-J assessments were performed by certified neuropsychologists [21, 22].　 The apolipoprotein E (ApoE) haplotype was determined by polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) method as described previously [23]. All participants were treated with lecanemab and followed up for a minimum of six months after the initiation of lecanemab treatment. During this period, brain MRI scans were performed as pre-treatment screening, followed by MRI at one month, three months, and six months after treatment initiation to assess the presence or absence of ARIA.

CSF samples were obtained through lumbar puncture at the L3/L4 or L4/L5 interspace and collected in polypropylene vials. These samples were promptly transported to SRL Inc. (Tokyo) without freezing, where concentrations of Aβ1–40, Aβ1–42, and tau phosphorylated at threonine 181 (p-tau181) were measured using the LUMIPULSE G600II platform [24, 25].

Participants underwent initial pre-treatment screening MRI within one month of CSF collection under standardized conditions.

All MRI scans were performed using a 3.0T Philips Elition scanner (Philips Healthcare, Best, The Netherlands) with a 32-channel head coil. The imaging protocol included the following sequences:

CSO-PVS were evaluated on T2-weighted MRIs obtained at the initial pre-treatment screening, based on the criteria established by Charidimou et al [14]. Small sharply delineated structures of (or close to) CSF intensity measuring < 3 mm and following the course of perforating or medullary vessels on the slice where the hand knobs of the primary motor cortex were observed were visually counted by two trained observers (TK, KA) independently. An independent adjudicator (KFM) decided on cases where the two examiners' decisions were inconsistent. TK: Neurologist with 26 years of experience in diagnostic imaging for cognitive disorders, KA: board-certified radiologist with 24 years of experience. KFM: Neurologist with 17 years of experience. All observers were blinded to the clinical data.

Data are presented as mean ± standard deviation (SD). Mann-Whitney U test was used for two-group comparisons using GraphPad Prism 10. Fisher’s exact test was used to evaluate the statistical association between two categorical variables using GraphPad Prism 10. Kappa statistics and intraclass correlation coefficient analyses were conducted using SPSS version 23. Statistical significance was set at two-tailed P < 0.05.

In accordance with the dichotomization as described above, participants were categorized into low-degree and high-degree CSO-PVS groups. The clinical characteristics of the two groups are presented in Table 2. There were no significant differences between the two groups regarding sex, age, dementia diagnosis (MCI-due to AD or mild AD), history of hypertension, dyslipidemia, diabetes, atrial fibrillation, stroke, or ApoE genotype. MMSE-J scores were significantly lower in the low-degree CSO-PVS group than in the high-degree group. There were four cases with ARIA E/H during 6-months observation period, of which one case was symptomatic. Among the four cases with ARIA E/H, three belonged to the high-degree CSO-PVS group, while one was in the low-degree CSO-PVS group. The case with symptomatic ARIA belonged to the low-degree CSO-PVS group (as reported previously [26]). There was no significant difference in the incidence of ARIA E/H between the two groups.

Figure 3 presents the concentrations of Aβ40, Aβ42, and p-tau181 in CSF. CSF Aβ40 levels were significantly lower in the high-degree CSO-PVS group compared to the low-degree group (P = 0.0235). CSF Aβ42 levels also tended to be lower in the high-degree group, although the difference narrowly missed statistical significance (P = 0.0502). CSF p-tau181 levels were significantly lower in the high-degree group than in the low-degree group (P = 0.0264). The CSF Aβ42/40 ratio did not differ significantly between the high- and low-degree groups (P = 0.5453). However, as hypothesized, the product of Aβ40 and Aβ42 (Aβ40 × Aβ42) was significantly lower in the high-degree group (P = 0.0264).

In this study, participants with high-degree of CSO-PVS exhibited lower CSF Aβ40 levels. Although the difference did not reach statistical significance, Aβ42 levels also tended to be lower in the high-degree CSO-PVS group. Importantly, this study targeted patients with mild AD or MCI due to AD who were eligible for lecanemab treatment, thus including only individuals with minimal CAA-related hemorrhagic findings and few enlarged CSO-PVS. Indeed, only one case exhibited 21 or more enlarged CSO-PVS, the threshold defined by the Boston criteria version 2.0 [8]. Even though only a few participants exhibited a high degree of CSO-PVS in this cohort, it was noteworthy that further reductions in CSF Aβ42 and Aβ40 levels—biomarkers of vascular amyloid deposition—were still observed in accordance with the relative severity of the CSO-PVS burden. Previous studies have reported significantly reduced Aβ40 and Aβ42 levels in patients with sporadic CAA meeting the modified Boston criteria and hereditary Dutch CAA compared to control groups [9, 13] [27]. Additionally, the severity of CSO-PVS has been shown to correlate with the extent of CAA-related hemorrhages and white matter lesions [14]. The findings of the current study align with the results predicted from those previous reports. In particular, considering the fact that our subjects exhibited less CAA-related hemorrhages, the current results suggest the implication that CSO-PVS dilation and associated reductions in CSF Aβ42 and Aβ40 might precede the development of CAA-related cerebral hemorrhages. This interpretation is consistent with prior research in hereditary Dutch CAA, which demonstrated that reduction in CSF Aβ42 and Aβ40 as well as dilation of CSO-PVS were evident even in asymptomatic carriers[13, 28]. Notably, the product of Aβ42 and Aβ40 concentrations (Aβ42 × Aβ40) was also significantly lower in the high-degree CSO-PVS group. This composite metric may reflect the cumulative loss of both longer (Aβ42) and shorter (Aβ40) amyloid species into vascular deposits and has the advantage of mitigating potential variability from individual peptide measurements. Although not commonly used, such multiplicative indices have been proposed as exploratory biomarkers in previous fluid biomarker studies [15]. To the best of our knowledge, there are few studies that have conducted a direct comparison between CSO-PVS grading and reductions in Aβ40 and Aβ42 levels. Our findings imply that dilated CSO-PVS may serve as a potential diagnostic marker reflecting vascular amyloid deposition for candidates for anti-amyloid therapy who exhibited minimal or no CAA-related hemorrhages.

Interestingly, CSF p-tau181 levels were lower in the high-degree CSO-PVS group. This may reflect impaired glymphatic or perivascular clearance vascular amyloid, which hinders the transport of phosphorylated tau into the CSF [29]. Furthermore, given that vascular amyloid deposition per se has limited neurotoxicity to the brain parenchyma unless accompanied by hemorrhage[7], it is conceivable that the increase in p-tau181—a biomarker reflecting both tau phosphorylation and the presence of plaque—was attenuated in the high-degree CSO-PVS group [30]. This latter hypothesis is in line with our finding that the high-degree CSO-PVS group demonstrated significantly higher MMSE-J scores.

Finally, we acknowledge that the small sample size represents a significant limitation of this study, potentially reducing the statistical power to detect associations between CSO-PVS severity and CSF biomarkers. Future research should focus on developing objective, automated methodologies for CSO-PVS grading, minimizing reliance on visual assessments. Moreover, large-scale, multicenter studies are necessary to further elucidate the relationship between CSF biomarkers for amyloids and CSO-PVS findings.

Among individuals with mild AD or MCI due to AD characterized by reduced CSF Aβ42/40 ratio and minimal or no CAA-related hemorrhages, CSF Aβ40 levels were significantly lower and CSF Aβ42 levels tended to be lower in those with a high degree of CSO-PVS compared with those with a low degree of CSO-PVS. These findings suggest that dilated CSO-PVS may serve as a potential surrogate marker of vascular amyloid deposition in patients undergoing anti-amyloid therapy.

Figure 1 Hypothetical Framework of the Study

Figure 2

Figure 3

The CSF biomarkers, including Aβ42 (a), Aβ40 (b), p-tau181 (c), the Aβ42/40 ratio (d), and the product of Aβ40 and Aβ42 concentrations (Aβ40 × Aβ42) (e), were compared between the low-degree and high-degree groups of CSO-PVS. Statistically significant differences between the two groups are indicated by asterisks. Bars indicate median values.