Comparative methods for quantifying plasma biomarkers in Alzheimer's disease: Implications for the next frontier in cerebral amyloid angiopathy diagnostics

Abstract Plasma amyloid beta (Aβ) and tau are emerging as accessible biomarkers for Alzheimer's disease (AD). However, many assays exist with variable test performances, highlighting the need for a comparative assessment to identify the most valid assays for future use in AD and to apply to other settings in which the same biomarkers may be useful, namely, cerebral amyloid angiopathy (CAA). CAA is a progressive cerebrovascular disease characterized by deposition of Aβ40 and Aβ42 in cortical and leptomeningeal vessels. Novel immunotherapies for AD can induce amyloid‐related imaging abnormalities resembling CAA‐related inflammation. Few studies have evaluated plasma biomarkers in CAA. Identifying a CAA signature could facilitate diagnosis, prognosis, and a safer selection of patients with AD for emerging immunotherapies. This review evaluates studies that compare the diagnostic test performance of plasma biomarker techniques in AD and cerebrovascular and plasma biomarker profiles of CAA; it also discusses novel hypotheses and future avenues for plasma biomarker research in CAA.

imperfect.Clinical diagnosis does not correlate well with pathologic diagnosis on post mortem assessment, which may be related to the clinical heterogeneity in cognitive presentation and overlap with other neurodegenerative and neurovascular conditions impairing other cognitive domains. 1,6Over the past two decades, AD diagnostics have rapidly expanded to include neuroimaging biomarkers derived from amyloid and tau positron emission tomography (PET); cerebrospinal fluid (CSF) Aβ and tau analyses; and, more recently, plasma-based biomarkers of p-tau, total tau (t-tau), and Aβ. 1 The most validated CSF biomarkers in AD are a decreased Aβ 42/40 ratio, as well as increased t-tau and p-tau181, with accumulating evidence for p-tau231, and p-tau217. 1,7,8Furthermore, in one study, CSF p-tau/Aβ 42 predicted the conversion from mild cognitive impairment (MCI) to AD dementia with 82.9% sensitivity and 90% specificity. 9However, the high cost and limited availability of PET, and the invasiveness of CSF collection, makes plasma biomarkers a more appealing approach for large-scale screening efforts. 10 with AD, a definitive diagnosis of CAA is only possible with neuropathology.3][4][5] The 2022 Boston criteria 2.0 for CAA now also include supportive criteria of enlarged perivascular spaces in the centrum semi-ovale and white-matter hyperintensities in a multi-spot pattern-the non-hemorrhagic neuroimaging features of CAA. 2 However, the diagnostic specificity remains limited.The current Boston criteria 2.0, for distinguishing probable compared to no CAA, have a sensitivity of 80% (95% confidence interval [CI]: 71%-88%) and specificity of 82% (95% CI: 62%-94%), though when distinguishing probable or possible compared to no CAA these criteria have a sensitivity of 90% (CI: 82%-95%) and specificity of 59% (95% CI: 39%-78%). 2The typical neuroimaging features of CAA on magnetic resonance imaging (MRI) according to Boston criteria 2.0 are depicted in Figure 1.
Plasma biomarkers could have a role in CAA diagnosis.Enlarged perivascular spaces may reflect impaired glymphatic clearance across the neurovascular unit. 11Impaired perivascular drainage of Aβ is hypothesized to have a major role in the pathogenesis of CAA. 3,12As CAA is a disorder of cerebral small vessels and results from the aberrant clearance and subsequent accumulation of amyloid surrounding and within small vessels, it is biologically plausible that measuring the spillage of these poorly cleared waste products into plasma, across the neurovascular unit, may serve as a useful approach to directly investigate and diagnose CAA.Furthermore, given the proposed selective dysfunction in the neurovascular glymphatic interface in CAA, we might expect a different plasma biomarker profile in CAA compared to AD without CAA.Identifying plasma biomarkers of CAA could facilitate accurate and early diagnosis, which may be the next natural advancement to further increase the diagnostic sensitivity and specificity of the Boston criteria.However, CSF and plasma amyloid have been relatively less studied in CAA compared to AD.

RESEARCH IN CONTEXT
1. Systematic review: Pubmed/MEDLINE searches identified studies which (a) compared diagnostic test performances of plasma amyloid beta (Aβ) and tau assays in Alzheimer's disease (AD), and (b) characterized spinal fluid and plasma biomarkers of cerebral amyloid angiopathy (CAA).

Interpretation:
A comparative assessment of plasma biomarker assay performance identified several plasma biomarker assays with variable test performance in AD.
While mass spectrometry displayed the best diagnostic test performance for plasma Aβ quantification, good to excellent performance was noted across most assays for tau.Furthermore, in CAA, few have evaluated spinal fluid biomarkers and there is a scarcity of evidence evaluating plasma biomarkers.The pathophysiologic spectrum of AD and CAA, their overlap, and summary of amyloid and tau synthesis are summarized in Figure 2 along with hypothesized corresponding changes in plasma biomarkers across this disease spectrum.The amyloidogenic pathway through βand γsecretases generate aberrant Aβ isoforms.Aβ 42 is the least soluble, and has a higher propensity to aggregate and deposit as parenchymal plaques in AD. 1,3,13 The shorter Aβ 40 isoform is more soluble and is preferentially cleared along perivascular pathways. 1,3,13In those with CAA, Aβ 40 deposits within and causes damage to cerebral small blood vessels. 2,12,14While the traditional understanding of the pathobiology of AD has centered on injury mediated by amyloid plaque and tau tangles, there is increasing appreciation for the pathophysiologic role of a reactive astrocytosis as a pathophysiologic pathway interacting with amyloid pathology. 15,16One potential biomarker of reactive astrocytosis is plasma glial fibrillary acidic protein (GFAP), which is elevated in AD and is a marker of early cerebral amyloidosis, but not tauopathy. 15,16While few studies have examined GFAP in CAA, a recent study found no association between GFAP and CAA pathology. 16Additionally, neurofilaments are cytoskeletal scaffolding components inside neurons that may serve as sensitive markers of neuronal injury. 17,18Neurofilament light chain (NfL) is a marker of axonal injury and is elevated in AD and across several neurodegenerative disorders of the central nervous system. 17,18e recent rapid growth in plasma biomarker studies in AD has been enabled by the development of accessible, reliable, and more accurate techniques.Historically, enzyme linked immunosorbent assay (ELISA) methodologies were implemented, which have limited capacity to detect biomarkers in the small quantities present in plasma. 19th the advent of single molecule protein detection (Simoa) and immunoprecipitation mass spectrometry (IP-MS) assays, the ability to detect these biomarkers in plasma has improved.1][22][23][24] The plasma Aβ 42 /40 ratio, p-tau181, p-tau217, p-tau231, and t-tau have emerged as potentially sensitive and specific measures to discriminate AD. 20,21,[25][26][27][28][29][30][31][32] As exciting as this rapidly expanding plasma biomarker space is, a critical review of the various methodologies now available is warranted to identify the biomarkers and techniques with the best accuracy and reliability for future consideration.The first objective of this narrative review is to provide a focused overview of recent studies that have directly compared the diagnostic performance of different assays used to quantify plasma biomarkers in AD in reference to standard diagnostic tests (PET, CSF biomarkers, or pathology) or in reference to conversion to clinical AD dementia or compared to healthy cognitively normal controls.Identifying studies that have directly compared methods within the same group of participants and using the same diagnostic reference standard for those participants facilitates a more homogenous and accurate comparison of plasma diagnostic test performance characteristics, as opposed to heterogenous comparisons of different methods in different participants across different studies.
The results of this first objective will identify the most valid biomarker assays for future use in AD and identify valid assays to then be applied to future CAA research.Building on our first objective, our second aim is to review CSF and plasma biomarker studies that have been published in CAA to identify knowledge gaps and opportunities to apply the most valid assays identified from our first objective to CAA.Furthermore, we will discuss the diagnostic and potential treatment implications of these findings for AD and CAA.

LITERATURE REVIEW
In accordance with methodological standards for narrative reviews, we followed the Scale for the Assessment of Narrative Review Articles (SANRA; see Table S1 in supporting information). 33 I G U R E 2 Pathophysiologic spectrum of Alzheimer's disease and cerebral amyloid angiopathy and hypothesized plasma biomarker alterations across the spectrum.APP is metabolized by β-secretase and γ-secretase to produce Aβ isoforms between 27 and 43 amino acids in length. 1,14When APP is metabolized abnormally two pathologic isoforms are generated: Aβ 40 and Aβ 42 .Aβ 42 is the least soluble, and has a higher propensity to aggregate and deposit in the brain parenchyma as neuritic plaques in AD. 1,3,13 The shorter Aβ 40 isoform is more soluble and is preferentially cleared along perivascular pathways. 1,3,13In some persons, beta-amyloid deposits within small cerebral arteries and arterioles to cause CAA. 2,12,14.The deposition of beta-amyloid weakens cerebral vessels, resulting in loss of smooth muscle and subsequent hemorrhaging.While Aβ 42 primarily deposits in plaques, it also deposits within cortical and leptomeningeal small vessels in CAA, but the ratio of Aβ 40 : Aβ 42 is higher in cerebral blood vessels than in the parenchymal amyloid plaques. 3Furthermore, in AD, intracellular tau tangles also accumulate and are responsible for neuronal injury.

RESULTS
For our literature search pertaining to our main objective, 1185 publications were identified for title and abstract review.5][36][37][38][39][40][41][42]   The Simoa Amyblood method, mentioned above, uses a more specific antibody, which quantifies the full length of Aβ 42 and Aβ 40 and not fragments, and which ultimately improves the specificity of this assay. 38,43While one study found that the Simoa Amyblood method could discern AD on Amyloid PET with an AUC of 0.79 (95% CI: 0.73-0.85), 38 MSD may have arisen due to differences in the antibodies used in each platform. 40oot et al. 41 directly compared two p-tau217 Simoa assays from Eli Lilly and Janssen Research and Development and demonstrated excellent AUCs in two independent cohorts (see Table 2).Finally, perhaps the highest quality evidence found for head-to-head comparisons of various plasma p-tau techniques was by Janelidze et al. 42 examining p-tau181, p-tau217, and p-tau231 across 10 plasma assays including IP-MS-and Simoa-based methods in 135 patients with MCI followed for 4.9 years. 42The results from Janelidze et al. 42 are summarized in Table 2 and  A few studies to date have evaluated the diagnostic test performance of GFAP in AD.In one study, plasma GFAP concentration was elevated in PET amyloid positive patients compared to those who were PET amyloid negative, 15 AUC = 0.761, which was superior to CSF GFAP AUC = 0.694 (95% CIs not reported). 15This is consistent with another study in which the diagnostic test performance of plasma GFAP was greater than CSF GFAP and the change in the magnitude of plasma GFAP was greater than changes in CSF GFAP in AD. 48This study also demonstrated elevations in plasma GFAP in pre-clinical AD and in those with symptomatic AD.Similarly, CSF amyloid positive status is associated with plasma GFAP (AUC = 0.72, 95% CI: 0.70-0.74)

Plasma NfL and GFAP in AD
and NfL (AUC = 0.60, 95% CI: 0.57-0.64),although not as strongly as plasma Aβ 42/40 (AUC = 0.79, 95% CI: 0.76-0.82)and plasma p-tau 217 (AUC = 0.82, 95% CI: 0.80-0.85). 49Plasma GFAP has been reported to predict incident clinical diagnoses of AD at 0 to 17 years with an AUC = 0.729 (95% CI: 0.682-0.776),which was better than plasma p-tau181 (AUC = 0.610, 95% CI: 0.556-0.776)and plasma NfL (AUC = 0.676, 95% CI: 0.629-0.724). 50ere are some studies comparing NfL between AD and other neurodegenerative or neurological conditions that can cause later life cognitive decline, an important comparison as these are conditions that would have to be differentiated from AD in clinical practice.Plasma NfL is elevated across multiple neurodegenerative conditions including AD, amnestic MCI, and dementia with Lewy bodies (DLB), but is especially elevated in frontotemporal dementia and amyotrophic lateral sclerosis. 51Plasma GFAP has mostly been evaluated in AD, though one study also reported elevations in plasma GFAP in DLB 52 and a recent study reported elevations in plasma GFAP in the cerebellar subtype of multi-system atrophy. 53More studies are needed to directly compare plasma GFAP across neurodegenerative disorders.[56][57][58][59][60] However, it is worth exploring whether levels of NfL or GFAP, in combination with more specific markers of disease processes, can improve diagnostic accuracy.

STUDIES OF CORE CSF AND PLASMA BIOMARKERS IN CAA
Our MEDLINE/Pubmed search yielded 425 titles and abstracts.The articles extracted and reviewed in full are detailed below for CSF and plasma biomarkers of CAA.

CSF
Overall, two meta-analyses and six case control studies were identified, evaluating the core CSF biomarkers of Aβ 42 , Aβ 40 , p-tau, and t-tau in relation to Boston criteria probable or possible CAA.After the time frame of our literature search, we found one additional study recently published, comparing cohorts of CAA, AD, MCI related to AD, and normal controls.These nine studies are summarized in Table 3. [61][62][63][64][65][66][67][68] Some of these studies are limited by small sample sizes.
A meta-analysis by Charidimou et al. 68 pooled results from three studies, 62,65,66 with a total of 59 patients with CAA, 158 patients with AD, and 94 HC.These studies are also independently summarized in Table 3. 62,65,66 68 With respect to the discrimination between CAA and AD, CSF Aβ 40 RoM = 0.76 (95% CI: 0.69-0.83)was the most discriminative, while no difference in CSF Aβ 42 was noted. 68nce this meta-analysis by Charidimou et al. 68 was published in 2018, three additional case-control studies have been published, 63,64,67 one of which also includes an updated metaanalysis. 67A larger case-control study of modified Boston criteria probable CAA patients is reported in Grangeon et al. 63 In this study authors used ELISA to evaluate CSF Aβ 42 , Aβ 40 , p-tau, and t-tau between 63 patients with CAA, 27 patients with AD, and 21 controls. 63erall, 85% of patients who had probable CAA according to revised Boston criteria had abnormal CSF biomarkers. 63Patients with CAA and AD had similar levels of Aβ 42 , but overall, CAA had lower levels of CSF t-tau, p-tau, and Aβ 40 .That said, three CSF profiles were identified among those with CAA: (1) an AD-like profile with elevated tau and p-tau levels (50.8%), (2) an isolated reduction in Aβ 42 (34.9%),and (3) normal Aβ 42 (14.3%). 63Across all three CAA profiles, however, a reduction in Aβ 40 was uniformly a distinguishing characteristic compared to AD cases. 635][66] In this most recent meta-analysis, summarized in Figure 5
The plasma Aβ 42/40 ratio was decreased in CAA compared to HC, driven by a higher plasma Aβ 40 , in CAA compared to HC. 70 The mean Aβ 40 detected in serum of CAA patients was 5.92 ± 1.41 pg/mg and in normal controls it was 4.13 ± 1.28pg/mg, a difference significant at P = 0.0002. 70Another study using multiplexing technology with detecting microspheres found that plasma Aβ 42 , truncated fragments of Aβ 40 , and full-length Aβ 40 were elevated in patients with CAA-related intracerebral hemorrhage compared to HC. 71 Using a highly sensitive dielectrophoretic driven biosensor, Kim et al. 72  1.0-1.0). 72The findings from these studies support the hypothesis that amyloid deposition in the vasculature causes spillover into the systematic circulation.
In contrast, a study using Simoa to measure plasma Aβ in presymptomatic Dutch type hereditary CAA found that both Aβ 40 and Aβ 42 were reduced in mutation carriers (n = 9) compared to noncarriers (NC; n = 8). 73This was the only identified study in which CAA cases were confirmed by genetic or neuropathological data, and thus the diagnosis of CAA in this study was probably more accurate.These data also raise the possibility that plasma Aβ  67 framework for CAA has been published and it would be interesting for future studies to evaluate whether there might be variations in plasma Aβ concentrations according to stage. 74nally, another study examined the association of plasma Aβ 40 concentrations, measured by ELISA, with white matter hyperintensities in patients with AD, MCI, and CAA. 75In this study, plasma Aβ 42 concentrations were lower in AD and MCI cohorts than in CAA, but Aβ 40 concentration was similar across the three groups. 75However, there was no control group so the study could not conclude whether Aβ 40 and Aβ 42 differed between CAA and persons without MCI or AD dementia.

CAA-related inflammation
CAA-related inflammation (CAA-ri) is an autoimmune disorder characterized by the production of antibodies against Aβ deposited within the walls of cortical and leptomeningeal vessels. 76,777][78] While CAA-ri is associated with similar neuroimaging characteristics depicted in Figure 1 such as microbleeds, macro-hemorrhages, and superficial siderosis, some of the differentiating neuroimaging hallmarks that can occur with CAA-ri include: (1) acute and confluent T2 fluid attenuated inversion recovery hyperintensities affecting subcortical U-fibers manifesting as a result of acute cerebral vasogenic edema, (2) mass effect, and [80] CAA-ri is characterized by perivascular inflammation with or without frank vasculitis.When vasculitis is present it is has also been termed Aβ related angiitis (ABRA). 81[84][85] CAA-ri, therefore, has been proposed as a naturally occurring model of amyloid immunotherapy-induced ARIA. 77,86 is important for clinicians to recognize CAA-ri as many patients appear to respond to immunosuppressive agents including glucocorticoids. 77,86,87Advances in understanding the evidence-based management of CAA-ri may help inform the appropriate approach to managing ARIA-E and ARIA-H as well, which will become more common as lecanemab and donanemab enter clinical practice.In the largest registry of 113 participants with CAA-ri treated with immunosuppression, 70.3% of participants had clinical recovery and 45.1% had radiographic recovery at 3 months, with an increase in radiologic recovery to 77.4% at 1 year. 77CAA-ri recurred in 38.3% in the subsequent 24 months, and recurrence was more likely if oral corticosteroid tapering occurred rapidly. 77finitive diagnosis of CAA-ri requires biopsy with neuropathological confirmation.However, it is increasingly common for patients to be diagnosed without biopsy, using validated neuroradiological criteria. 76erging evidence suggests that fluid markers may prove to be useful in the diagnosis and monitoring of CAA-ri.Using an ultrasensitive ELISA assay with magnetic beads, elevated anti-amyloid autoantibody levels have been detected in CSF during the active phases of CAAri, which decline to near normal levels during the remission phases of CAA-ri in response to immunosuppression. 86Furthermore, the concentration of CSF anti-amyloid autoantibodies was positively correlated with the concentration of CSF Aβ 40 and Aβ 42 in this study, although this correlation was stronger with the soluble isoform Aβ 40 . 86all case series using amyloid PET suggests that, in CAA-ri, there is reduced amyloid deposition in regions with the greatest CAA-ri associated cerebral edema-perhaps suggesting that the neuroinflammatory response facilitates the removal of cerebral amyloid. 88In another study, elevations in CSF anti-amyloid autoantibodies were noted in those with CAA-ri but not in those patients with either AD or MCI who had multiple cerebral microbleeds or confluent white matter hyperintensities, suggesting that the presence of anti-amyloid autoantibodies in CSF may be a specific marker of CAA-ri and not merely a consequence of asymptomatic concomitant CAA. 89ile CSF anti-amyloid antibodies may be a biomarker of CAA-ri, there are no reported studies, to the best of our knowledge, evaluating whether specific plasma anti-amyloid antibodies may distinguish CAA-ri from CAA without inflammation and/or AD.One hypothesis, supported by one cohort study, is that CSF anti-amyloid antibodies in CAA-ri are intrathecally produced; 87 therefore, they may not necessarily be associated with plasma anti-amyloid antibodies.However, naturally occurring plasma anti-amyloid antibodies have been described in AD. 90,91 While several studies have demonstrated a reduction in the concentration of naturally occurring plasma antiamyloid antibodies in AD compared to HC, [90][91][92][93][94] others have found elevations. 95,96The clinical significance of these naturally occurring anti-amyloid antibodies is uncertain, but they have been proposed to be protective, 97 though in another study neither their presence nor concentration were associated with dementia risk. 98These naturally occurring amyloid antibodies in plasma may be related to cerebral amyloidosis, but no study to date has demonstrated an association with plasma antibodies and CAA-ri.

DISCUSSION
Overall, this narrative review revealed several studies within the same participants directly comparing the diagnostic test performance of various plasma Aβ and tau assays in reference to PET or CSF reference standards for AD.We did not find any comparative studies with autopsy used as the diagnostic reference standard.Recognizing this limitation, novel modifications have attempted to increase Simoa's specificity.For instance the Simoa Amyblood method uses a more specific antibody quantifying the full length of Aβ 42 and Aβ 40 . 35,38,43The results of our review do identify significant variability in the performance of Simoa assays for Aβ, suggesting the need for further optimization of Simoa technology for Aβ quantification specifically or potentially the use of IP-MS or LC-MS methods for Aβ quantification.
Another important consideration when evaluating different studies reporting variable AUCs is that differences in PET reference standards may have resulted in lower agreement of some Aβ 42/40 assays compared to previously reported AUCs. 25,28Florbetapir PET may have greater variability than PiB. 112However, an earlier study using the IP-MS technique from Washington University used PiB-PET and found that plasma Aβ 42/40 had an AUC of 0.887, 113 which is comparable to AUCs reported in Table 1 with the same IP-MS technique from Washington University but using flutemetamol 34 and florbetapir 35  In this study, p-tau231 displayed greater association with tau PET and amyloid PET, compared to p-tau181. 36Interestingly this study also noted that the ratio of the Aβ 40/42 ratio to p-tau231 predicted a faster decline in immediate and delayed memory domains in participants at risk of AD. 36 In AD, neurofibrillary tangles are associated with the severity of AD dementia more than Aβ plaques, 115 and it would be prudent to examine whether the same holds true in cases of CAA without AD.A combination of plasma p-tau217, Aβ 42/40 , and NfL together with basic demographic data, predicted subsequent development of AD dementia with AUC = 0.82 (95% CI: 0.77-0.91) in an analysis of the BioFINDER study, thereby demonstrating the utility of plasma biomarkers evaluated combined with clinical factors to increase our diagnostic discriminative ability. 116Future studies examining the diagnostic capabilities of plasma biomarkers in CAA similarly should assess test performances individually for each biomarker, but also in combination in multivariable modeling.
In addition to the discriminative ability of plasma Aβ in discerning the presence of AD, plasma tau species have also emerged as valuable discriminative biomarkers.Among 1402 participants across three cohorts, p-tau217 discriminated AD from other neurodegenerative diseases with higher accuracy than other plasma and neuroimaging biomarkers with an AUC 0.96 (95% CI: 0.93−0.98)using the Eli Lilly MSD immunoassay. 21In another study p-tau181 and p-tau231 measured via MSD immunoassay could discriminate AD from non-AD neurodegenerative disease with AUCs > 0.90 and both were associated with amyloid and tau PET positivity. 117Furthermore, consistent across several different IP-MS-, MSD-, and Simoa-based technologies p-tau217, p-tau181, and p-tau231 had good discriminative ability for predicting conversion from MCI to AD dementia and for discriminating AD based on CSF Aβ. 117 The results in Table 2 and Figure 4, especially highlighting the study by Janelidze et al., 42 61 Overall, the consequences of parenchymal and vascular trapping of amyloid isoforms (depicted in Figure 2) are reductions in CSF Aβ 40 and Aβ 42 . 61,64The fact that Aβ 40 was especially different between CAA and non-CAA groups further supports the vascular propensity of Aβ 40 and why it may be a hallmark biomarker in CAA. 61,64Additionally, this study supports the hypothesis that elevations in t-tau and p-tau are greatest in AD, in which there is greater neuronal loss related to concomitant tauopathy, compared to CAA. 61 Therefore, the combination of Aβ 40 , Aβ 42 , p-tau, and t-tau likely will help facilitate the differentiation between AD and CAA.Further, this combination may also help identify cases of concomitant CAA and AD.In these cases, it is possible that there will be elevations in tau isoforms that mirror those levels seen in AD and reductions in Aβ 40 that mirror those levels seen in CAA.While insufficient evidence exists to conclude whether the same biomarker profile might occur in plasma, these findings would largely be supportive of the pathophysiologic model in Figure 2 and does invite future research to evaluate whether a potential plasma profile may be similar to the CSF profile identified.
There are limitations to the studies that have evaluated plasma or CSF biomarkers of CAA that are worth mentioning.First, most have done so in reference to possible and probable CAA and not definite, neuropathologically confirmed, CAA according to Boston criteria.Furthermore, across the studies evaluating CSF and plasma biomarkers of CAA, some only included those with probable CAA, while others included those with possible CAA, thereby adding to the heterogeneity in diagnostic certainty.Other studies also include those with acute ICH, adding further potential heterogeneity according to disease state or activity. 61,68,118It is uncertain whether plasma and/or CSF biomarkers may be altered in the setting of acute ICH.Studies evaluating plasma biomarkers in CAA have been especially limited by small sample sizes and variability in the methodologies used to quantify plasma biomarkers in CAA and, therefore, it is hard to draw definitive conclusions from this small body of literature.This calls for future studies in larger cohorts using robust plasma-based methods to quantify and evaluate the core plasma biomarkers of CAA.
We have reached the point at which reliable tau and Aβ plasma assays have been developed in AD, and now efforts ought to examine the potential diagnostic utility and validation of these biomarkers and assays in CAA.Not only could plasma biomarkers increase the diagnostic sensitivity and specificity of the Boston criteria 2.0 for CAA even further, but there are clinical scenarios which may benefit from their implementation.
Having summarized several studies that directly compare different assays in the same patients with the same outcome measure (amyloid PET or CSF amyloid for instance) this review facilitates the identification of methods with the highest validity and reliability to quantify plasma biomarkers of AD in the future.The identified robust plasma methods in AD will help inform the selection of the best assays to evaluate potential biomarkers in CAA.Evaluating plasma biomarkers in CAA offers the potential to improve our clinical diagnostic approach.While there have been significant advances in the diagnosis and tailored treatments of acute ischemic stroke, both diagnostics and treatments for hemorrhagic stroke have lagged.In those with lobar hemorrhage, it can often be difficult to discern whether hemorrhage is due to CAA especially when neuroimaging biomarkers of CAA are not present. 2,3asma biomarkers of CAA might improve our diagnostic accuracy in patients with ICH and could inform the future risk of hemorrhage.One study found that plasma full-length Aβ 42 , truncated fragments, and fulllength Aβ 40 were elevated in patients with CAA-related ICH compared to HC. 71 The clinical utility in measuring plasma biomarkers in CAA is also evidenced by another study, in which NfL predicted an increased risk of ICH. 119The validation of plasma biomarkers in CAA will also facilitate the evaluation of novel hypotheses.We postulate whether there may be a stage-dependent alteration in plasma Aβ 40 given its propensity to accumulate within cerebral small vessels in CAA.Some of the studies we identified, bearing in mind their small sample sizes, identified increases in plasma Aβ 40 in CAA-related hemorrhage. 72 (postulated in Figure 2).This hypothesis needs to be further evaluated as identifying such a rise in plasma Aβ 40 could have wide implications for informing prognosis and the risk of hemorrhage in patients with CAA.Furthermore, the use of anticoagulation for other medical indications is not uncommon in elderly cohorts in which CAA can co-occur.
Plasma biomarkers of CAA, if validated and found to be predictive of hemorrhage, therefore, might also have potential implications to safely guide anticoagulation decision making as well in the future.
Evaluating potential plasma biomarkers of CAA also has relevance for the novel and emerging immunotherapies in AD, which facilitate the removal of cerebral Aβ. 82 Therapies such as aducanumab 82 and lecanemab 120 for early AD facilitate Aβ clearance from the brain, but they can cause ARIA-E and ARIA-H, including microhemorrhage and leptomeningeal superficial siderosis. 82Those with apolipoprotein E (APOEA) ε4 positivity are at particular risk of ARIA. 82yloid-targeting monoclonal antibodies, therefore, can induce an often subclinical neuroimaging pattern that resembles CAA and CAA-ri.
Identifying a plasma biomarker profile of CAA-ri could facilitate diagnosis of CAA-ri and provide a more feasible, cost-effective means of monitoring response to immunotherapy that could potentially guide the intensity of therapy.As CAA-ri is rare, it would be prudent to support international collaboratives that store plasma samples from patients for future exploration of new markers.Future studies could use the most accurate and reliable plasma methods identified by this narrative review to explore potential differences in plasma Aβ -42/40 , t-tau, p-tau isoforms, GFAP, and NfL in those with CAA-ri, CAA without inflammation, AD, and controls.One might hypothesize that, given the extent of neuroinflammation in CAA-ri, a marker of reactive astrocytopathy, GFAP, may be particularly elevated in those with CAA-ri.
Additionally, studies are needed to determine whether plasma antiamyloid antibodies can be detected that correlate with the increased levels of CSF anti-amyloid antibodies that are known to occur during CAA-ri with active inflammation.
Overall, as immunotherapies for AD become increasingly used, there must be diagnostic prudence to identify those with concomitant CAA. 85[85] A plasma biomarker profile specific to CAA could facilitate the safe selection of patients with AD for disease-modifying therapies.Furthermore, identifying plasma biomarkers of CAA might also help monitor the biological response to immunotherapies in AD, if vascular and parenchymal amyloid can be discriminated. 120The potential for plasma biomarkers to be used to monitor therapies is exhibited in the TRAILBLAZER-ALZ randomized, double blind, placebo-controlled clinical trial with donanemab, in which plasma levels of p-tau217 and GFAP were significantly reduced in the donanemab treatment arm compared to placebo. 121In contrast to the development of therapeutics for AD, the therapeutic arena for CAA has been limited, perhaps out of concern that targeting Aβ in CAA might exacerbate the risk of hemorrhage.However, new apoE immunotherapies in animal models have been demonstrated to reduce amyloid pathology while preserving cerebrovascular integrity in CAA. 122To continue to advance the therapeutic front in CAA, we must also understand the unique plasma biomarker profile in CAA, whether this profile might predict risk of future hemorrhage, and how biomarkers might have a role in monitoring a biological response to emerging targeted therapies.

CONCLUSION
This Reduced CSF t-tau and p-tau in CAA compared to AD may also be helpful. 61,63Overall these studies invite expansion of research explor- This is a novel hypothesis that needs formal evaluation.Finally, our narrative review also places the utility of identifying a plasma biomarker profile of CAA in the context of emerging immunotherapies for AD.
Given that immunotherapies, like lecanemab, can induce ARIA-H and ARIA-E and neuroimaging features that resemble CAA and CAA-ri, identifying a sensitive and specific plasma biomarker profile of CAA could help prognosticate, monitor, and safely select those with AD for immunotherapies in the future.

3 .
Future directions: This review identifies a need to characterize and validate CAA plasma biomarkers.Such markers could increase the feasibility and accuracy of CAA diagnosis in clinical practice, and potentially improve the prediction of risk for CAA-related amyloid-related imaging abnormalities in patients undergoing immunotherapies for AD.We propose novel hypotheses including a stagedependent alteration in plasma Aβ 40 in CAA and potential implications for increasing the diagnostic specificity of the Boston criteria and informing hemorrhage risk.

F I G U R E 1
Neuroimaging characteristics of cerebral amyloid angiopathy.(A) SWI depicting multiple cortical cerebral microbleeds.(B) GRE depicting focal superficial siderosis in the bilateral frontal lobes.(C) GRE depicting focal superficial siderosis in the right posterior parietal lobe.(D) GRE depicting diffuse cortical superficial siderosis.(E) GRE depicting a focus of chronic hematoma cavity with hemosiderin staining in the left frontal lobe.(F) SWI depicting a large left frontal intracerebral hemorrhage with mild mass effect and midline shift.(G) T2-weighted image demonstrating multiple enlarged perivascular spaces in the centrum semiovale of both cerebral hemispheres.(H) FLAIR sequence depicting white matter hyperintensities in a multi-spot pattern throughout the centrum semiovale in bilateral hemispheres.FLAIR, fluid attenuated inversion recovery; GRE, T2*-weighted gradient recalled echo; SWI, susceptibility weighted image.

Figure 4
and demonstrate good to excellent AUCs for most IP-MS-and Simoa-based methods at detecting AD based on CSF Aβ 40/42 and in discriminating those who progress from MCI to AD.
quantified Aβ 40 and Aβ 42 in 25 persons without AD and 19 with AD.These participants were further classified as probable CAA according to modified Boston criteria.Among those with probable CAA, Aβ 40 was elevated compared to those without CAA and Aβ 40 discriminated probable CAA from AD with an AUC = 1.0 (95% CI: 40 concentrations could vary by CAA disease stage, with low Aβ 40 in the pre-symptomatic CAA phase followed by high Aβ 40 in the symptomatic phase when there is vessel wall breakdown and hemorrhage.A recent disease stage F I G U R E 5 Diagnostic test performance of cerebrospinal fluid biomarkers in cerebral amyloid angiopathy.Figure generated from data extracted from Margraf et al.
ing potential plasma biomarkers in CAA.Aβ 40 has a greater propensity for depositing in the small vessels of the cerebral vasculature and, hence, its greater specificity as a biomarker for CAA over AD in CSF studies to date.However, the limited literature examining plasma biomarkers in CAA have inconsistently identified both elevations and reductions in plasma Aβ 40 , highlighting an important need for further research using robust assays to clarify these discrepancies.We hypothesize that while changes in plasma Aβ 40 levels might be congruous with CSF Aβ 40 , there could also be a stage-dependent alteration in plasma Aβ 40 .Reductions in plasma Aβ 40 may be congruous with reductions in CSF Aβ 40 during earlier pre-hemorrhagic stages of CAA, while elevations in plasma Aβ 40 might occur in later hemorrhagic stages of CAA.

Table 2 . 38-42 3.1 Plasma Aβ 42/40 in AD and MCI
Summary of studies directly comparing plasma Aβ biomarker quantification techniques in AD.
39oforms in relation to CSF Aβ 42/40 and amyloid PET reference standards are depicted in Table2and Figure4.With respect to tau-based assays, De Meyer et al.38did not directly compare t-tau individually, but rather compared an ELISA to Simoa Aβ 42 /t-tau ratio.In CSF it has been reported that Aβ 42 /t-tau is a better predictor of progression in AD than the Aβ 42/40 ratio alone.38Asdisplayed in Table2, the AUC for both methods for the Aβ 42 /t-tau ratio was identical.In another study, Bayoumy et al.39compared six different 40cha et al., in an ADNI cohort, found an AUC of 0.66 (95% CI: 0.56-0.76)forthismethod.35Anotherrecentstudyused a modified Simoa technique to quantify plasma Aβ 42/40 from plasma neu-The main AUC point estimates and corresponding 95% CIs extracted from studies evaluating the performance of various plasma tau techniques across tau ticipants and compared two p-tau Simoa methods and a Meso Scale Discovery (MSD) Platform (Eli Lilly) against both Pittsburgh compound B (PiB) amyloid PET and F 18 flortaucipir tau PET.The MSD works like an ELISA and uses electrochemiluminescence.40Mielke et al.40evaluated p-tau231 and p-tau217, but they did not compare two different quantification methods head to head for these two biomarkers.However, Mielke et al.40did compare two different p-tau181 methods and found that with respect to amyloid PET in patients with MCI, Simoa Quanterix had an AUC of 0.82 (95% CI: 0.63-1.00),while Eli Lilly MSD had an AUC of 0.78 (95% CI: 0.58-0.98).With respect to tau PET, in patients with MCI Simoa Quanterix had an AUC of 0.88 (95% CI: 0.66-1.09)while Eli Lilly MSD had an AUC of 0.86 (95% CI: 0.67-1.04).The very small sample size of MCI patients in this study likely contributed to the wide confidence limits.Differences between Simoa Quanterix and Eli Lilly 45te with most studies measuring the performance of these biomarkers in reference to a clinical diagnosis of AD and fewer evaluating these in reference to CSF amyloid or PET markers.Studies of NfL suggest that it may have some value in discriminating AD from controls without neurodegenerative diseases.In one study using data from ADNI, a Quanterix Simoa assay was used to evaluate plasma NfL and discriminated the AD dementia group from normal controls with an AUC = 0.87, while CSF Aβ 42 had an AUC = 0.88 and CSF t-tau had an AUC = 0.90 (95% CIs were not reported).45Another This narrative review did not identify any comparative analyses of different plasma methods to quantify NfL or GFAP biomarkers.While both are elevated in CSF and plasma in AD, there has been variable diagnostic test performance of NfL and GFAP reported in the literature to Summary of studies in CAA evaluating the core CSF biomarkers (Aβ 42 , Aβ 40 , p-tau, and t-tau in relation to Boston criteria probable or possible CAA).Aβ 42 /Aβ 40 , AUC = 0.69, (95% CI: 0.63-0.76)⚬ Aβ 40 , AUC = 0.73 (95% CI: 0.67-0.79)⚬ Aβ 42 , AUC = 0.54 (95% CI: 0.47-0.61)⚬ ptau181, AUC = 0.76 (95% CI: 0.69-0.82)⚬ t-tau, AUC = 0.71 (95% CI: 0.64-0.77) may decrease in the early pre-symptomatic phase and then increase in the symptomatic hemorrhage phase.Studies to further evaluate these hypotheses are needed.In one of the first studies using ELISA to quantify plasma Aβ 40 and Aβ 42 , between 25 patients with probable or definitive CAA related intracerebral hemorrhages (ICH) and 21 patients with hypertensive TA B L E 3 ⚬ Aβ 42 /Aβ 40 AUC = 0.62 (95% CI: 0.43-0.81)⚬ Aβ 42 AUC = 0.79 (95% CI: 0.66-0.92)⚬ Aβ 40 AUC = 0.69 (95% CI: 0.51-0.87)⚬ tau AUC = 0.60 (95% CI: 0.43-0.77)⚬ p-tau AUC = 0.60 (95% CI: 0.44-0.76)• CAA and AD: • CAA and controls ratio of means: ⚬ Aβ 42 0.49 (95% CI: 0.38-0.64),P < 0.0001 ⚬ Aβ 40 0.70 (95% CI: 0.63-0.78),P < 0.0001 ⚬ t-tau 1.54 (95% CI: 1.15-2.07)P = 0.004 ⚬ p-tau 1.24 (95% CI: 0.99-1.54),P = 0.06 • CAA and AD ratio of means: 46,[102][103][104][105][106][107][108][109][110][111] x-40 .10TheAUCsreportedforAβ42/40 for Simoa Quanterix assays have a much greater variability in the ability to differentiate AD groups based on CSF Aβ 42/40 , or amyloid PET, ranging from 0.58 to 0.92.46,[102][103][104][105][106][107][108][109][110][111]For this reason IP-MS methods, as evident in Table1, are likely superior to Simoa methods for the quantification of Aβ in plasma. therefore 11437s reference standards.Furthermore, several studies have noted that the majority of discordant Aβ 42/40 plasma and PET cases are Aβ 42/40plasma positive and amyloid-PET negative.28,37Anotherimportantconsiderationwhen evaluating the test performance of methods in reference to PET is that it is possible that abnormal changes in plasma Aβ 42/40 might precede the threshold at which Aβ cortical PET positivity is reached.This is consistent with prior work demonstrating that CSF Aβ changes are detected before Aβ PET positivity.114Therefore,one important knowledge gap and direction for future studies is to evaluate how the diagnostic accuracy of plasma biomarkers may vary according to the stage of AD.Furthermore, there is a continuing We did not find comparative studies evaluating the sensitivity and specificity of plasma biomarker assays in CAA.Furthermore, to date, no study has examined the sensitivity and specificity of plasma Aβ 40 , Aβ 42 , Aβ 40/42 , and p-tau isoforms together in CAA.
61e largest case-control study to date evaluating CSF biomarkers in CAA61demonstrates largely consistent findings with prior work supporting that amyloid, particularly CSF Aβ 40 , is primarily reduced in CAA and that both CSF Aβ 40 and Aβ 42 have good to excellent diagnostic discriminative abilities.
61rrative review provides a critical evaluation of the diagnostic test performances of several plasma Aβ 42 , Aβ 40 , and tau isoform assays in AD.Comparative studies consistently demonstrate that Simoa, MSD, and mass spectrometry methodologies have good to excellent test performance for plasma tau evaluation, while mass spectrometry methods consistently perform better for the quantification of plasma Aβ 42 and Aβ 40 isoforms.The summation and critique of knowledge presented in this review from such comparative studies and the summation of this body of knowledge into a schematic depiction of pathophysiologic underpinnings of CAA and the corresponding potential alterations in plasma biomarkers in Figure 2 is novel and will facilitate the selection of the most accurate and reliable methods for future research as well as the translation of these methods to investigate CAA.Our review identifies a substantive knowledge gap in the assessment of potential plasma biomarkers in CAA and calls for dedicated research in this arena.We hypothesize that the identification of a novel plasma biomarker profile of CAA may additively facilitate its diagnosis in combination with Boston criteria and could, in the future, be predictive of prognosis and hemorrhage risk in CAA.The CSF biomarker profile of CAA is better characterized than plasma.The recently published, largest case-control study to date supports that CSF Aβ 42 and Aβ 40 isoforms have good to excellent discriminative ability for CAA.61