Contradiction and Falsification Atlas

A useful hypothesis page contains the strongest case for the claim, the strongest challenge, alternative explanations, a falsifier, and a decisive experiment. Confidence is not a vote count.

H001 | Perivascular clearance failure is an initiating driver of sporadic CAA

For: Anatomical plausibility, vascular amyloid distribution, hereditary/experimental models, and spatial PVS-pathology associations

Against: Human flow direction and temporal onset are not directly established; clearance impairment may be secondary to vessel injury

Alternative: Production, vessel affinity, age-related matrix change, pulsatility, and BBB dysfunction

Falsifier: Normal direct clearance before and during early vascular amyloid accumulation despite sensitive measurement

Decisive experiment: Longitudinal compartment-specific human transport study beginning before imaging disease

H002 | Vascular physiological dysfunction precedes structural cSVD injury

For: Longitudinal CVR predicts later WMH/PVS progression; staged CAA frameworks place physiology before tissue injury

Against: Reverse causation and shared vascular risks remain possible; clinical outcomes were not predicted at one year in MSS3

Alternative: Existing microscopic tissue injury could impair measured reactivity

Falsifier: Repeated normal physiology immediately before incident lesions in adequately powered cohorts

Decisive experiment: Repeated physiology and structural imaging with cross-lagged and intervention analyses

H003 | BBB leakage is an early causal amplifier of CAA vessel injury

For: Leakage and inflammation increase with histologic vessel injury and fit ARIA/CAA-ri biology

Against: Human evidence is mainly cross-sectional and leakage can follow vessel damage

Alternative: Inflammation, ischemia, and amyloid toxicity independently generate permeability

Falsifier: No leakage before progression when measured serially with sensitive methods

Decisive experiment: Serial DCE-MRI/fluid markers with spatial pathology and treatment perturbation

H004 | A strictly lobar hemorrhagic pattern is sufficiently specific for CAA

For: Strong pathology-linked enrichment and established Boston criteria performance in symptomatic cohorts

Against: Nonamyloid arteriolosclerosis can produce some lobar/juxtacortical microbleeds; performance falls in other spectra

Alternative: Trauma, embolic lesions, vascular malformations, treatment effects, and mixed disease

Falsifier: High rate of non-CAA pathology among rigorously localized strictly lobar lesions

Decisive experiment: Lesion-matched MRI-histology plus external diagnostic validation

H005 | CSO-PVS are an in-vivo marker of regional cortical CAA burden

For: Spatial MRI-pathology association with overlying cortical amyloid and smooth-muscle loss

Against: PVS are common with aging and multiple SVD mechanisms; rating and causal interpretation vary

Alternative: Atrophy, altered interstitial fluid, venous factors, and other small-vessel pathology

Falsifier: No regional association in larger blinded pathology cohorts after confounding control

Decisive experiment: Multiregion MRI-autopsy mapping with reproducible automated PVS quantification

H006 | A deep/perforator MRI pattern can diagnose brain arteriolosclerosis

For: Biological plausibility and candidate associations of deep lesions, WMH, infarcts, diffusion, and vascular risks

Against: No accepted pathology-validated criteria; lesions are downstream, mixed, and sequence-dependent

Alternative: CAA, embolism, diabetes, aging, AD, inflammation, and other arteriopathies

Falsifier: Poor external calibration against standardized multiregion pathology

Decisive experiment: Large derivation plus independent locked MRI-autopsy validation

H007 | Arteriolosclerosis independently causes cognitive decline

For: Autopsy associations persist after adjustment in several aging cohorts

Against: Residual confounding, measurement error, mediation by infarcts/WMH, and cohort reuse

Alternative: ADNC, LATE, hippocampal sclerosis, frailty, and vascular risks

Falsifier: No association after robust measurement-error and causal-pathway modeling across independent cohorts

Decisive experiment: Replicated longitudinal autopsy mediation analysis with harmonized pathology

H008 | ARTS is a valid in-vivo biomarker of arteriolosclerosis

For: Pathology-linked development, reported AUC, translation to in-vivo MRI, and repeatability work

Against: Limited independent pathology validation, selected cohorts, and uncertain incremental utility

Alternative: ARTS may reflect general aging, WMH/diffusion injury, or vascular risk burden

Falsifier: Failure of locked calibration/discrimination in an independent MRI-autopsy cohort

Decisive experiment: Independent preregistered external pathology validation against strong baselines

For: Overlapping ARIA-E/ARIA-H imaging, vascular amyloid, permeability, inflammation, and occasional clinicopathologic resemblance

Against: Distinct triggers, treatment exposure, natural history, ascertainment, and incomplete direct comparative data

Alternative: Shared downstream injury from different initiating mechanisms

Falsifier: Consistently nonoverlapping immune, imaging, and pathology signatures in direct comparisons

Decisive experiment: Harmonized prospective CAA-ri versus ARIA multi-omic/imaging/pathology study

H010 | Regional amyloid PET can distinguish CAA from AD

For: Meta-analysis finds lower global CAA uptake than AD and relatively higher occipital-to-global uptake

Against: Most reference diagnoses are clinical; AD co-pathology and tracer heterogeneity blur compartments

Alternative: Disease stage, tracer affinity, atrophy, and mixed parenchymal/vascular amyloid

Falsifier: No patient-level separation against compartment-specific autopsy pathology

Decisive experiment: Prospective multi-tracer PET-to-autopsy threshold validation

H011 | A fluid biomarker panel can distinguish CAA from AD and arteriolosclerosis

For: Disease-related amyloid, neuronal, inflammatory, endothelial, and pericyte signals are detectable

Against: Substantial construct overlap, assay heterogeneity, and little pathology-linked external validation

Alternative: Systemic vascular disease, renal function, age, AD, and nonspecific neurodegeneration

Falsifier: No incremental accuracy beyond clinical/MRI models in representative pathology cohorts

Decisive experiment: Locked multisite plasma/CSF panel tested against mixed autopsy pathology

H012 | Mixed CAA, arteriolosclerosis, and neurodegenerative pathology interact rather than merely add

For: Frequent co-occurrence, shared vascular/amyloid pathways, spatial vulnerability, and heterogeneous phenotypes

Against: Interaction estimates are underpowered, scale-dependent, and vulnerable to measurement error

Alternative: Independent additive effects plus age-related co-occurrence

Falsifier: Consistently additive models with good calibration across independent cohorts and scales

Decisive experiment: Large replicated multiregion pathology-imaging cohort with prespecified interaction scales