Debates, Hypotheses, and Open Questions

How to read this chapter

Each section distinguishes the hypothesis, observations that support it, observations that constrain it, and a study that could discriminate among explanations. "Evidence against" often means evidence against a simple or universal version, not proof that the whole mechanism is false.

1. Is CAA fundamentally a perivascular clearance disease?

Hypothesis

Soluble Aβ normally exits brain along perivascular pathways. Aging or vascular dysfunction reduces clearance; Aβ accumulates in vessel walls; deposition further impairs drainage, producing a feed-forward loop.

Supporting observations

• Vascular deposition follows compartments plausibly involved in interstitial-fluid drainage.
• Experimental tracer work demonstrates periarterial/perivascular movement.
• Severe centrum semiovale PVS are associated with CAA in imaging cohorts and contribute to Boston v2.0.
• Perosa et al. spatially linked dilated white-matter PVS to Aβ accumulation and smooth-muscle loss in overlying cortical vessels.
• Hereditary CAA data suggest molecular/physiologic abnormalities long before hemorrhage.

Constraints and alternatives

• MRI-visible PVS are a structural endpoint and do not measure flow direction or Aβ flux.
• PVS enlargement may be caused by amyloid, atrophy, inflammation, or shared aging processes.
• IPAD and glymphatic models differ in routes and driving forces; the field sometimes combines them too loosely.
• Anesthesia, sleep, posture, arterial pulsation, and species strongly affect experimental flow.
• Parenchymal and vascular amyloid can share production and clearance factors without one route being dominant.

Discriminating study

A longitudinal human study in presymptomatic hereditary CAA combining dynamic molecular tracer methods, quantitative PVS morphology, vascular reactivity, CSF/plasma Aβ kinetics, arterial pulsatility, sleep/posture measures, and repeated MRI, with preregistered temporal predictions. A mechanistic model should predict which variable changes first and where.

2. Does vascular dysfunction precede visible tissue injury?

Hypothesis

Impaired cerebrovascular reactivity, autoregulation, BBB function, and microvascular flow occur before WMH, microinfarcts, or hemorrhage and define a treatable stage.

Supporting observations

• Dutch-type CAA studies support early reactivity abnormalities.
• The Koemans framework orders physiologic dysfunction before non-hemorrhagic and hemorrhagic injury.
• Diffusion and network measures often correlate with cognition beyond visible lesion counts.

Constraints

• Most sporadic studies are cross-sectional.
• Vascular reactivity depends on baseline CO2, medications, hematocrit, neural activation, vascular anatomy, and analysis.
• Tissue injury can itself impair measured reactivity, creating reverse causation.

Discriminating study

Repeated physiology and tissue-imaging measurements in initially lesion-free, high-risk participants, analyzed with within-person change and negative-control regions. Demonstrate that physiologic change predicts subsequent spatially matched injury after controlling for baseline occult damage.

3. Is BBB leakage a cause, consequence, or amplifier?

Hypotheses

1. Barrier failure is an early upstream lesion allowing plasma proteins, inflammation, and edema to damage tissue.
2. Barrier leakage occurs after vessel-wall injury and is mainly a marker.
3. Both are correct in subtype- and stage-dependent feedback loops.

Evidence

Kozberg et al. observed leakage and smooth-muscle loss in mildly amyloid-affected vessels and leakage in remodeled vessels, while later vessel stages showed more perivascular inflammation. This spatial pathology supports a staged model but comes from a small autopsy series and cannot establish within-person temporal sequence.

Discriminating study

Register in vivo dynamic contrast-enhanced MRI and fluid markers to ex vivo whole-hemisphere MRI and serial histology. Sample vessels across regions prospectively, not only visible lesions, and quantify amyloid, junction proteins, fibrinogen, smooth muscle, matrix remodeling, immune cells, and microhemorrhage.

4. Why can a ruptured CAA vessel contain little amyloid at the rupture site?

Competing explanations

• local immune or proteolytic removal of Aβ precedes rupture but leaves a weakened wall;
• fibrinoid remodeling physically replaces the amyloid-bearing segment;
• the apparent depletion is a sampling/sectioning problem;
• hemorrhage destroys or displaces antigen;
• adjacent rather than directly sampled vessel segments contain the causal amyloid;
• some lobar microbleeds have a non-CAA cause.

What would decide the issue

Three-dimensional serial reconstruction of entire culprit vessels, with spatially resolved Aβ species, smooth muscle, collagen/basement membrane, fibrin, iron, complement, proteases, and immune cells, compared with nonruptured vessels from the same brain and arteriolosclerosis controls.

5. Are strictly lobar microbleeds specific to CAA?

Conventional model

Strictly lobar distribution strongly favors CAA; deep distribution favors arteriolosclerosis. This is diagnostically useful and underlies Boston criteria.

Challenge

Perosa et al. found true lobar CMBs caused by arteriolosclerosis in three pathology-negative, clinically false-positive CAA cases. Lesions tended toward juxtacortical/subcortical white matter rather than cortical ribbon.

Balanced conclusion

Anatomic distribution is a high-value probabilistic biomarker, not a deterministic lesion label. Patient-level diagnostic specificity can be high even when lesion-level specificity is imperfect. Higher-resolution localization relative to cortical ribbon and penetrating-vessel anatomy may improve etiologic inference.

Needed evidence

Large prospective MRI-autopsy datasets with lesion-by-lesion coregistration, standardized SWI/QSM, whole-vessel serial histology, and controls spanning mixed CAA/B-ASC. Report both patient-level and lesion-level accuracy.

6. Can WMH patterns distinguish CAA from arteriolosclerosis?

Proposed pattern

Posterior-predominant or multispot subcortical WMH may support CAA; confluent periventricular/deep WMH and deep lesions may support non-amyloid arteriopathy.

Problem

WMH are a final common pathway of edema, demyelination, axonal loss, rarefaction, gliosis, and incomplete infarction. Age, blood pressure, AD, CAA, B-ASC, venous disease, and technical segmentation all influence them.

The multispot pattern contributes to Boston v2.0 in combination with a lobar hemorrhagic lesion, but added diagnostic performance was smaller than that of severe CSO-PVS in the derivation work. A 2025 pilot linked multispot count to amyloid PET burden in 21 participants; the small sample and parenchymal/vascular PET ambiguity require caution.

Research direction

Move from a single pattern label to spatially explicit lesion morphology, rim/perfusion/permeability/diffusion features, vascular territories, and pathology-linked training. Validate against both CAA and B-ASC severity rather than against clinical labels alone.

7. Are enlarged PVS a clearance biomarker, severity marker, or epiphenomenon?

Possibilities

• direct structural consequence of impaired fluid clearance;
• marker of vessel stiffening and smooth-muscle loss;
• consequence of atrophy or tissue loss;
• route-dependent response to CAA in centrum semiovale and B-ASC in basal ganglia;
• nonspecific aging feature with disease-modified distribution.

Key methodological issue

Most studies count visible spaces on one slice or use thresholded segmentations. They do not measure flow, vessel identity, connectivity, or whether the PVS surrounds an artery or vein.

Needed tools

High-resolution 3D vessel/PVS imaging, arterial-versus-venous labeling, fluid-dynamic measurements, and tissue validation along the same penetrating vessel.

8. Is brain arteriolosclerosis one disease or a family of lesions?

Unitary view

Age, hypertension, diabetes, and metabolic/inflammatory stress converge on hyaline wall thickening, smooth-muscle loss, and luminal narrowing.

Heterogeneity view

The same semiquantitative score may combine fibrosis, muscular remodeling, basement-membrane duplication, leakage, mineralization, inflammatory change, fibrinoid injury, and region-specific processes with different causes and consequences.

Why it matters

A heterogeneous pathology label weakens biomarker specificity and can hide treatment-responsive subtypes. A classifier trained on a binary or coarse autopsy label inherits that ambiguity.

Discriminating study

Unsupervised and hypothesis-driven spatial pathology across multiple regions using vessel morphometry, matrix proteomics, cell phenotyping, and clinical exposures. Replicate clusters, relate them to MRI/ARTS, and test differential cognitive and stroke outcomes.

9. Does arteriolosclerosis cause hippocampal sclerosis and LATE-NC?

Evidence for a relationship

Multiregional arteriolosclerosis has been associated with hippocampal sclerosis of aging in large autopsy datasets, and B-ASC often coexists with LATE-NC. The medial temporal lobe is vulnerable to vascular and proteinopathic injury.

Alternative explanations

• common aging or genetic susceptibility;
• systemic vascular/metabolic exposure affects both;
• selection/collider bias in autopsy cohorts;
• neurodegeneration alters vascular demand or structure;
• regional tissue loss changes vessel-density measurements;
• unmeasured co-pathology mediates the association.

Needed evidence

Longitudinal antemortem vascular biomarkers plus regional MRI, followed by standardized LATE/hippocampal pathology. Use mediation models only after establishing temporal order and explicitly test common-cause models.

10. Is CAA independent of Alzheimer disease or part of one amyloid continuum?

Continuum arguments

Shared Aβ production, APOE susceptibility, age dependence, and frequent coexistence imply linked clearance and deposition biology.

Distinct-disease arguments

Different Aβ isoform enrichment, vascular versus parenchymal compartment, clinical syndromes, hereditary CAA without typical AD pathology, and cases of AD without severe CAA support separable processes.

Better formulation

CAA and AD are partially coupled compartment-specific amyloid disorders. The coupling strength likely varies by genotype, disease stage, Aβ species, vascular physiology, and clearance pathway. Analyses should model both burdens rather than forcing a binary choice.

11. Is neuroinflammation protective or destructive in CAA?

Protective possibilities

Microglia, macrophages, antibodies, and complement may clear vascular Aβ and contain injury.

Destructive possibilities

The same response may increase protease activity, barrier leakage, edema, smooth-muscle loss, and rupture. CAA-ri and ARIA show that immune-mediated vascular amyloid mobilization can be clinically harmful.

Central question

Which immune programs, in which vascular compartments and disease stages, promote safe clearance versus destructive remodeling?

Decisive approach

Spatial single-cell and proteomic analysis of vessels across CAA grade and rupture state, paired with longitudinal imaging in anti-Aβ-treated and untreated cohorts. Treat treatment-associated ARIA as a perturbation experiment but account for dose, antibody, APOE, baseline CMB/cSS, and AD severity.

12. Does intensive vascular-risk control prevent cSVD cognitive decline?

Blood-pressure trials show that intensive control can slow WMH accumulation in selected populations, and vascular health is a rational intervention target. Remaining uncertainties include the optimal target by age/frailty, whether established CAA behaves differently, competing ischemic and hemorrhagic risks, and how much cognitive benefit is mediated by cSVD rather than other pathways.

The research error is to infer mechanism solely because a risk-factor intervention changes a downstream marker. Mediation requires measurement of the proposed pathway and attention to treatment-induced competing effects.

13. Should anti-amyloid therapy be viewed as a CAA experiment?

Anti-Aβ antibodies mobilize amyloid and can cause ARIA-E/H, with risk influenced by APOE, baseline hemorrhagic lesions, treatment, and likely vascular amyloid. This offers evidence that vascular amyloid handling and inflammation are mechanistically connected.

However, treatment-associated ARIA is not a pure experiment on spontaneous CAA: participants have AD, drug exposure is nonphysiologic, MRI surveillance alters detection, and trial exclusion criteria truncate the CAA spectrum. The most informative studies integrate baseline CAA markers, drug-specific exposure, longitudinal fluid/imaging measures, and pathology when available.

14. Do composite SVD scores clarify or obscure?

Value

They summarize burden, improve statistical power, reduce multiple testing, and can stratify prognosis.

Cost

They threshold continuous biology, weight lesions arbitrarily, mix causes and stages, and can hide opposing component effects. A 0-4 total SVD score and a 0-6 CAA-SVD score are not interchangeable.

Best practice

Report the composite and all components. Predefine missing-data handling. Test whether results persist with continuous measures, alternative weights, and latent-variable models.

15. A living hypothesis matrix

Update this matrix when new data change confidence, not merely when a new paper appears.