Skip to content

Nosology and diagnostic criteria

TL;DR — Vascular dementia criteria do not merely name the same patients differently; they encode different assumptions about memory, lesion burden, timing, and causal certainty (Sachdev 2014, PMID 24632990; O'Brien 2015, PMID 26595643). NINDS–AIREN prioritized specificity for research, VASCOG broadened the construct to mild and major vascular cognitive disorders, and VICCCS sought harmonized operational practice (Skrobot 2018, PMID 29055812). New VasCog-2-WSO criteria continue that convergence project (Sachdev 2025, PMID 40955506). None solves the central problem that vascular lesions are common in older adults and Alzheimer copathology is often present (Schneider 2007, PMID 17568013). A defensible diagnosis therefore reports syndrome severity, vascular phenotype, evidence for linkage, and competing pathology rather than a bare label.

What criteria must decide

Decision Competing choices Consequence
Cognitive threshold impairment on testing vs functional dependence Separates mild from major disorder
Required domains memory mandatory vs any affected domain Changes sensitivity to subcortical disease (Sachdev 2014, PMID 24632990)
Vascular evidence clinical stroke, imaging, or both Determines inclusion of covert disease
Causal linkage temporal, anatomical, severity-based Controls false attribution
Imaging minimum qualitative “relevant disease” vs explicit lesion patterns Changes reproducibility (Duering 2023, PMID 37236211)
Mixed disease exclusion, “possible,” or dual diagnosis Alters prevalence and trial populations
Certainty possible/probable/definite “Definite” often requires pathology

Criteria families

Framework Core architecture Strength Limitation
Hachinski Ischemic Score Clinical features separating “multi-infarct” from “primary degenerative” dementia Historically linked low CBF to the infarct group (Hachinski 1975, PMID 1164215) Score is not a modern imaging-based diagnosis
ADDTC Probable/possible ischaemic vascular dementia with clinical and imaging evidence Early operational research framework including mixed categories (Chui 1992, PMID 1549205) Narrower historical construct; variable implementation
NINDS–AIREN Dementia, cerebrovascular disease, and a relationship between them, stratified definite/probable/possible High specificity suited to trials; required temporal stroke–dementia linkage and imaging support (Román 1993, PMID 8094895) Can miss mild, covert, and slowly progressive disease (O'Brien 2015, PMID 26595643)
DSM-5 vascular NCD Mild/major neurocognitive disorder with vascular etiology Fits cross-etiology clinical taxonomy Imaging and lesion thresholds remain less granular
VASCOG Mild and major vascular cognitive disorders; memory not compulsory Cognitive phenotype better matches vascular syndromes Attribution still judgment-dependent (Sachdev 2014, PMID 24632990)
VICCCS Consensus classification plus harmonized cognitive/imaging protocols Improves comparability Consensus does not itself validate prognosis (Skrobot 2018, PMID 29055812)
VasCog-2-WSO Revised contemporary criteria Incorporates a decade of imaging/biomarker work; Delphi with 49–54 of 70 invited experts and WSO endorsement (Sachdev 2025, PMID 40955506) External performance needs prospective testing
NINDS–CSN protocols Harmonized 5/30/60-minute cognitive batteries and imaging standards Makes VCI research comparable (Hachinski 2006, PMID 16917086) Harmonization is not a diagnostic test

NINDS–AIREN required dementia, cerebrovascular disease, and a relationship between them, emphasizing infarct, haemorrhage, and hypoxic-ischemic heterogeneity and a temporal stroke–dementia link for a secure diagnosis (Román 1993, PMID 8094895). ADDTC operationalized ischaemic vascular dementia with neuroimaging and mixed categories (Chui 1992, PMID 1549205). VasCog-2-WSO added preclinical, mild, and major levels under VCID after three Delphi rounds (Sachdev 2025, PMID 40955506).

How badly do the criteria disagree? The validation numbers

"Criteria are not interchangeable" is usually stated as a caution. It has actually been measured, and the numbers are worse than the caution implies.

Agreement between criteria

Seven Alzheimer's Disease Diagnostic and Treatment Centers applied structured checklists for four vascular-dementia definitions to 25 standardized case vignettes. The frequency of a vascular-dementia diagnosis was highest with the modified Hachinski Ischemic Score and DSM-IV, intermediate with the original HIS and ADDTC, and lowest with NINDS–AIREN; interrater reliability ranged from κ = 0.30 for ADDTC to κ = 0.61 for the original HIS (Chui 2000, PMID 10681076). The oldest and least sophisticated instrument was the most reproducible — a durable finding, because reliability rewards simple checklists while validity rewards nuance, and the two goals pull in opposite directions.

Accuracy against autopsy

Study Sample Criteria Sensitivity Specificity
Gold 1997 (PMID 9305324) 113 autopsied demented patients, acute-care geriatric hospital ADDTC possible VaD 0.63 0.64
NINDS–AIREN possible VaD 0.58 0.80
Hachinski Ischemic Score 0.43 0.88
Gold 2002 (PMID 11772694) 89 autopsied patients (20 VaD, 23 mixed, 46 AD) ADDTC possible 0.70 0.78
NINDS–AIREN possible 0.55 0.84
DSM-IV 0.50 0.84
ADDTC probable 0.25 0.91
NINDS–AIREN probable 0.20 0.93
ICD-10 0.20 0.94
Bacchetta 2007 (PMID 16580095) 110 autopsy cases in the oldest-old (36 confirmed VaD) ADDTC / NINDS–AIREN / HIS possible 0.56–0.58 (all three) 0.74 / 0.73 / 0.66

Three conclusions follow directly and none has been superseded by a newer validation study.

  1. The "probable" tiers are close to unusable as case-finding instruments. ADDTC probable (sensitivity 0.25) and NINDS–AIREN probable (0.20) miss four in five autopsy-confirmed vascular dementias, and Gold and colleagues found no statistically significant relationship between neuropathological diagnosis and three of the four criteria sets tested — ICD-10, ADDTC probable, and NINDS–AIREN probable — concluding they "should be revised" (Gold 2002, PMID 11772694). Trials that recruit on probable VaD are therefore selecting an unrepresentative, high-specificity minority; this is the most likely single explanation for why VaD drug trials generalize so poorly (see treatment).
  2. What the criteria are actually good at is excluding Alzheimer disease, not identifying vascular disease. ADDTC correctly excluded 87% of pathological AD, NINDS–AIREN 91%, HIS 97% (Gold 1997, PMID 9305324). The failure mode is mixed dementia: 54% of mixed cases were misclassified as VaD by ADDTC, 29% by NINDS–AIREN, and 18% by HIS in 1997; in the 2002 series the proportion of mixed cases labelled vascular ran from 9% to 39% depending on criteria (Gold 1997, PMID 9305324; Gold 2002, PMID 11772694). Since mixed pathology is the modal finding at autopsy (Schneider 2007, PMID 17568013), the largest error source is the category the criteria handle worst.
  3. Accuracy degrades with age, and for a specific reason. In nonagenarians and centenarians, ADDTC sensitivity fell with age because 42% of pure vascular-dementia cases had no history of stroke at all; 30% of mixed cases were called VaD by both ADDTC and NINDS–AIREN and 45.9% by HIS (Bacchetta 2007, PMID 16580095). Criteria that hinge on a documented stroke event are structurally blind to the form of vascular dementia that dominates the oldest old — precisely the fastest-growing population.

The counterpoint, and it is a real one: these numbers come mostly from hospital autopsy series. The one population-based clinicopathological validation reaches a much more favourable verdict. Among 419 incident dementia cases in the Rochester Epidemiology Project (1985–1989) with 89 autopsies (median onset age 80), pathological diagnoses were Alzheimer disease in 51%, pure VaD in 13%, and combined AD+VaD in 12%; criteria requiring either a temporal stroke–dementia relationship or bilateral infarction in specified locations on imaging (the Mayo Clinic criteria) achieved sensitivity 0.75 and specificity 0.81 for pure VaD, positive likelihood ratio 3.9 (95% CI 2.2–6.7) (Knopman 2003, PMID 12707071). The design difference that produces the gap is instructive: Mayo made the temporal link and the imaging pattern alternative sufficient criteria rather than joint requirements, and the five pure-VaD cases missed were exactly those lacking the temporal relationship. Autopsy sensitivity of 0.20–0.25 is therefore not a property of clinical vascular diagnosis as such — it is a property of criteria that require converging evidence conjunctively, applied in referral populations enriched for mixed disease. Which estimate a reader should use depends on whether the target setting resembles an acute geriatric hospital or a community.

The modern criteria families (VASCOG, VICCCS, VasCog-2-WSO) were designed against exactly these failures, and none of them has yet been subjected to an equivalent autopsy-validation study. Their improvement over NINDS–AIREN is therefore argued from construct logic, not demonstrated against a pathological gold standard (Sachdev 2014, PMID 24632990; Skrobot 2018, PMID 29055812; Sachdev 2025, PMID 40955506).

The problem is upstream: "dementia" itself is not one definition

Before any vascular attribution rule applies, the dementia threshold has to be set, and the standard threshold definitions disagree with each other more than they disagree about causes. Applying six classification systems to the same 1,879 Canadian Study of Health and Aging participants aged ≥65 gave dementia prevalences from 3.1% (ICD-10) to 29.1% (DSM-III) — a tenfold range — and the systems identified different people: only 20 subjects met all six. Disagreement between DSM-III and ICD-10 was driven by long-term memory, executive function, social activities and symptom duration, and was largely unaffected by age, sex, education or institutionalization (Erkinjuntti 1997, PMID 9385127). The authors state the consequence bluntly: this affects "the right of many older persons to drive, make a will, and handle financial affairs".

The vascular layer multiplies the variance rather than resolving it. Applying five vascular-dementia criteria sets to 107 post-stroke patients who all met DSM-III dementia criteria classified 32.7% as VaD by NINDS-AIREN and 91.6% by DSM-IV, with ADDTC at 86.9% and DSM-III and ICD-10 both at 36.4%; only 31 of 107 met all five. Concordance was perfect between DSM-III and ICD-10 (κ=1.0), good between ICD-10 and NINDS-AIREN (κ=0.87), poor between most other pairs (ADDTC/DSM-IV κ=0.37), and the three discriminating requirements were focal neurological signs, unequal distribution of higher-cortical deficits, and imaging evidence of relevant cerebrovascular disease (Pohjasvaara 2000, PMID 11108755). A drug trial recruiting by NINDS-AIREN and an epidemiological survey using DSM-IV are not studying the same population, and neither is wrong.

Reliability within a single criteria set is better but not high. Four clinicians independently applying NINDS-AIREN to 42 dementia cases with clinical data and MRI achieved pairwise κ of 0.46 to 0.72 — moderate to substantial (Lopez 1994, PMID 8035923). That ceiling, reached under ideal conditions with standardized abstraction forms, bounds what any criteria-based prevalence estimate can mean.

Pooled across criteria and dementia types, however, reliability looks better than the vascular-specific literature suggests. A PROSPERO-registered systematic review of 22 studies (7,577 titles screened) found summary interrater κ of 0.66 (95% CI 0.53–0.78) for all-cause dementia by DSM-III-R, 0.71 (0.65–0.77) for Alzheimer disease by NINCDS-ADRDA, and 0.79 (0.70–0.87) for vascular dementia by ICD-10 — with Gwet's AC1/2 coefficients generally higher still (Cerullo 2021, PMID 33942363). Two readings are defensible and the literature has not chosen between them. Either vascular diagnosis is more reproducible than the Chui and Lopez studies implied, and those κ values reflect the difficulty of vignette-based and multi-criteria comparison rather than of routine practice; or ICD-10's high κ is bought with a definition so restrictive (3.1% dementia prevalence in the CSHA comparison below, autopsy sensitivity 0.20) that raters agree mainly because almost nobody qualifies. Reliability and validity are being measured on different axes, and only one study rated high quality on the GRRAS reporting standard (Cerullo 2021, PMID 33942363).

The modern criteria agree with each other, which is progress and also a limitation

Applying VASCOG, DSM-5 and VICCCS to 165 Sydney Stroke Study participants alongside NINDS-AIREN, ADDTC and DSM-IV, agreement among the three modern sets was excellent for both VaD and mild vascular cognitive disorder (κ = 0.83–1.0), while agreement between VASCOG and the older criteria for VaD was only κ = 0.47–0.63. VASCOG produced slightly lower mild-VCD rates than DSM-5 or VICCCS; VaD-based 10-year mortality prediction was similar across the three modern sets and higher than for the older ones; and prediction of incident dementia within 5 years from mild vascular cognitive disorder was slightly lower with VASCOG (Sachdev 2019, PMID 30927497). Convergence among criteria written by overlapping expert groups is weaker evidence of validity than convergence between independently derived ones would be — but the improvement in predictive validity over NINDS-AIREN and DSM-IV is real. VICCCS itself was produced by online Delphi consensus with a mean of 122 respondents from 27 countries (81% academic researchers) at a 67% consensus threshold, redefining VCI into mild and major forms with subtypes (Skrobot 2017, PMID 27960092).

Do the criteria select vascular pathology? One biomarker validation says yes

Criteria have almost never been validated against in-vivo pathology rather than against each other. In 186 memory-clinic patients in Singapore assessed with clinical, neuropsychological, MRI and ¹¹C-PiB amyloid PET, the AHA/ASA and DSM-5 criteria agreed almost perfectly (κ=1.00 for probable vascular MCI, probable VaD and possible VaD; κ=0.71 for possible vascular MCI) — and, more importantly, amyloid positivity fell sharply along the certainty gradient: 3.8% in probable vascular MCI and 15% in probable VaD, against 26.7% possible vascular MCI, 33.3% non-vascular MCI, 73.3% possible VaD and 76.2% non-VaD (P<0.001 under both criteria) (Folloso 2024, PMID 38651759). The "probable" tier is therefore doing what it claims — selecting people with little concomitant amyloid — which is the strongest defence of the current criteria available. It also confirms that the possible tier is a mixed-pathology category, not a weaker version of the same thing, and it does not address the low autopsy sensitivity of the probable tier discussed above: criteria can be highly specific and still miss most cases.

The alternative strategy: abandon the umbrella and define a homogeneous subtype

A second response to criteria failure — older than the harmonization projects and never formally retired — is to stop trying to make "vascular dementia" diagnosable and instead carve out one mechanistically coherent subtype for trials. Erkinjuntti and colleagues argued that definitional heterogeneity was itself "a factor in negative clinical trials" and proposed research criteria for subcortical ischaemic vascular dementia, modifying NINDS-AIREN to require small-vessel disease as the primary aetiology, lacunar infarcts and ischaemic white-matter lesions as the primary lesion type, subcortical location, and a subcortical clinical syndrome — merging the historical entities of Binswanger's disease and the lacunar state (Erkinjuntti 2000, PMID 10961414; Erkinjuntti 2002, PMID 11901245). The prediction attached to the proposal was explicit and testable: subcortical VaD should show a more predictable clinical picture, natural history, outcome, and treatment response.

That prediction was never properly tested, because the criteria were used to enrich trials rather than to compare trial yield against umbrella criteria. The strategy resurfaces unchanged in the FINESSE framework for small-vessel-disease trials two decades later (Markus 2022, PMID 35969390), which is a mark of how little the underlying problem moved. It is also in direct tension with VASCOG/VICCCS/VasCog-2-WSO, which broaden the construct rather than narrowing it: one programme optimizes for trial tractability, the other for clinical coverage, and no study has quantified what each costs the other.

The mild tier has a measured natural history, and it is not benign

Where mild vascular cognitive disorder sits on the trajectory matters for whether the category earns its place. In the Canadian Cohort Study of Cognitive Impairment and Related Dementias, of 146 clinic patients with cognitive impairment–not–dementia followed 2 years, 34% progressed to dementia and 14% reverted to not cognitively impaired; the vascular subtype (VCI-ND) converted at 40.0%, statistically indistinguishable from the pre-Alzheimer subtype at 41.0%, while psychiatric and not-otherwise-specified CIND had the highest reversion (20% and 30%) (Hsiung 2006, PMID 16966831). Two implications cut against common practice: a vascular mild-impairment label carries the same two-year dementia risk as a prodromal-Alzheimer label, and one in seven such diagnoses is transient — so a mild vascular cognitive disorder recorded at a single visit is a probabilistic statement about a fluctuating state, not a stage.

A layered diagnostic formulation

1. Establish acquired cognitive decline

History should establish change from previous ability, not merely a low score. Informant evidence is particularly important where processing speed, planning, medication management, or finances have changed but episodic-memory complaints are modest (Sachdev 2014, PMID 24632990).

2. Grade functional severity

Mild disorder preserves independence despite inefficiency or compensatory strategies. Major disorder interferes with independent instrumental or basic activities. Motor impairment after stroke can imitate cognitive dependence, so the functional judgment must isolate cognitive contribution where possible (Swartz 2025, PMID 39822128).

3. Characterize the cognitive phenotype

Executive and processing-speed impairment support a subcortical vascular pattern, but no single profile is sufficiently specific to establish etiology (Desmond 2004, PMID 15537510). Memory-predominant disease does not exclude vascular contribution, especially with strategic hippocampal/thalamic infarction or mixed Alzheimer pathology (Jellinger 2007, PMID 17324442).

4. Demonstrate vascular brain injury

MRI is preferred because FLAIR, diffusion, susceptibility, and structural sequences jointly reveal recent infarction, lacunes, white-matter hyperintensity, microbleeds, superficial siderosis, and atrophy. CT may show large infarcts, lacunes, and confluent disease but is less sensitive to small and haemorrhagic markers (Duering 2023, PMID 37236211; Razek 2021, PMID 33434866).

Link type Supporting pattern Main threat
Temporal decline begins within months of stroke or after recurrent events delirium, depression, pre-stroke decline
Anatomical lesion damages a cognitive network or strategic hub incidental lesion
Burden diffuse SVD sufficient to disrupt networks age-associated WMH without clear symptoms
Course steps coincide with vascular events gradual SVD and Alzheimer disease both progress
Negative evidence no better competing explanation incomplete biomarker workup

6. State competing and coexisting etiologies

Autopsy series show that forcing one primary label can erase clinically relevant copathology (Schneider 2007, PMID 17568013). Reports should state whether Alzheimer, Lewy-body, inflammatory, toxic-metabolic, sleep, psychiatric, or medication-related contributors remain plausible. See mixed pathology.

Imaging terminology and diagnostic discipline

STRIVE established common terms for recent small subcortical infarcts, lacunes, WMH, perivascular spaces, microbleeds, and brain atrophy; STRIVE-2 added advances in quantitative imaging and emphasized interpretation across disease processes (Wardlaw 2013, PMID 23867200; Duering 2023, PMID 37236211). A radiology phrase such as “chronic microvascular ischaemic change” is not itself a clinical diagnosis.

Report element Minimum useful detail
Infarcts number, location, vascular territory, age
WMH distribution and validated severity scale where possible
Lacunes count and topography
Microbleeds count and lobar/deep/mixed distribution
Superficial siderosis focal or disseminated
Perivascular spaces location and qualitative burden
Atrophy global and regional pattern
Acute lesion diffusion restriction and relation to symptoms

Criteria choice changes measured prevalence

Post-stroke studies illustrate the effect. Pendlebury and Rothwell found rates from 7.4% after first-ever stroke with pre-stroke dementia excluded to 41.3% after recurrent stroke with pre-stroke dementia included; study method and case mix explained 93% of variance (Pendlebury 2009, PMID 19782001). Barbay et al. estimated total post-stroke neurocognitive disorder at 53.4%, with cognitive-score threshold and recurrent-stroke proportion materially influencing estimates (Barbay 2018, PMID 30504699).

The lesson is not that one pooled number is “correct.” Every estimate should specify criteria, testing threshold, timing, stroke recurrence, pre-stroke exclusions, and setting.

Screening versus diagnosis

MoCA, MMSE, BMET, and abbreviated NINDS-CSN protocols are triage instruments. A systematic review found MoCA AUCs generally above 0.90 for vascular dementia versus controls and 0.87–0.93 for vascular mild cognitive impairment, but heterogeneity prevented meta-analysis (Ghafar 2019, PMID 31050033). In one 34-patient validation, full MoCA AUC was 0.950 and short MoCA AUC 0.936 versus MMSE 0.860; the small homogeneous sample limits transportability (Freitas 2012, PMID 22676901).

A positive screen establishes A positive screen does not establish
Need for fuller assessment Vascular etiology
Candidate affected domains Functional dependence
Baseline for serial comparison A universal disease cutoff
Need to review hearing/language/education Absence of delirium or depression

Minimum research reporting set

  • Criteria version and certainty category.
  • Premorbid function and education/language context.
  • Cognitive domains, instruments, norms, and impairment threshold.
  • Functional definition and informant source.
  • Imaging sequences, field strength, lesion definitions, and rating method.
  • Stroke timing, recurrence, and acute delirium.
  • Alzheimer and other biomarker availability.
  • Whether mixed cases were included, excluded, or separately analyzed.
  • Attrition, death as competing risk, and missing cognitive assessments.

These variables follow directly from the comparability failures that motivated VASCOG, VICCCS, and imaging harmonization (Sachdev 2014, PMID 24632990; Skrobot 2018, PMID 29055812; Duering 2023, PMID 37236211).

Open questions

  • Do VasCog-2-WSO categories predict decline and treatment response better than VASCOG and NINDS–AIREN? (Sachdev 2025, PMID 40955506)
  • What imaging burden is sufficient for causal attribution when Alzheimer biomarkers are positive? (Schneider 2007, PMID 17568013)
  • Can criteria remain globally usable without forcing MRI-dependent exclusion in low-resource settings? (Gorelick 2011, PMID 21778438)
  • Which screening thresholds are valid across language, education, stroke severity, and sensory disability? (Ghafar 2019, PMID 31050033)
  • Have VASCOG, VICCCS, or VasCog-2-WSO ever been validated against autopsy, as ADDTC, NINDS–AIREN, DSM-IV, ICD-10, and HIS were? (Gold 2002, PMID 11772694; Sachdev 2025, PMID 40955506)
  • If probable-VaD criteria have sensitivity 0.20–0.25, how much of the failure of VaD drug trials is attributable to recruiting on them? (Gold 2002, PMID 11772694)
  • How should criteria identify vascular dementia in the oldest old, where 42% of pathologically pure cases never had a clinical stroke? (Bacchetta 2007, PMID 16580095)
  • Can any criteria set exceed κ = 0.61 interrater reliability without reverting to a simple ischaemic checklist? (Chui 2000, PMID 10681076)
  • If six dementia definitions differ tenfold in prevalence and share only 20 of 1,879 cases, what does any single reported dementia prevalence mean? (Erkinjuntti 1997, PMID 9385127)
  • Should trials state the vascular criteria used as an inclusion parameter with the same prominence as the drug dose, given a 32.7%-to-91.6% classification range in one cohort? (Pohjasvaara 2000, PMID 11108755)
  • Is κ 0.46–0.72 an irreducible ceiling for clinical vascular attribution, or a criteria-design problem? (Lopez 1994, PMID 8035923)
  • Does agreement among VASCOG, DSM-5 and VICCCS reflect validity or shared authorship? (Sachdev 2019, PMID 30927497; Skrobot 2017, PMID 27960092)
  • Why do hospital-series validations give probable-VaD sensitivity 0.20–0.25 while a population-based series using disjunctive criteria gives 0.75 — is the difference case mix, criteria architecture, or both? (Gold 2002, PMID 11772694; Knopman 2003, PMID 12707071)
  • Is ICD-10's κ = 0.79 for vascular dementia evidence of reproducibility, or an artifact of a definition that almost nobody meets? (Cerullo 2021, PMID 33942363; Erkinjuntti 1997, PMID 9385127)
  • Has anyone tested Erkinjuntti's 2000 prediction that subcortical-VaD criteria yield more predictable treatment responses than umbrella criteria? (Erkinjuntti 2000, PMID 10961414; Markus 2022, PMID 35969390)
  • If 14% of clinic CIND diagnoses revert within 2 years, what confirmation interval should mild vascular cognitive disorder require before it is recorded as a diagnosis? (Hsiung 2006, PMID 16966831)
  • If "probable" VaD selects 15% amyloid positivity and "possible" VaD selects 73%, should the two tiers be treated as different diagnoses rather than different confidence levels? (Folloso 2024, PMID 38651759)

References

  1. Sachdev PS, et al. Diagnostic criteria for vascular cognitive disorders: a VASCOG statement. Alzheimer Dis Assoc Disord. 2014. PMID 24632990
  2. Skrobot OA, et al. Progress toward standardized diagnosis of vascular cognitive impairment. Alzheimers Dement. 2018. PMID 29055812
  3. VasCog-2-WSO Criteria Consortium. Revised Diagnostic Criteria for Vascular Cognitive Impairment and Dementia—The VasCog-2-WSO Criteria. JAMA Neurol. 2025. PMID 40955506
  4. O'Brien JT, Thomas A. Vascular dementia. Lancet. 2015. PMID 26595643
  5. Gorelick PB, et al. Vascular contributions to cognitive impairment and dementia. Stroke. 2011. PMID 21778438
  6. Schneider JA, et al. Mixed brain pathologies account for most dementia cases. Neurology. 2007. PMID 17568013
  7. Jellinger KA, Attems J. Neuropathological evaluation of mixed dementia. J Neurol Sci. 2007. PMID 17324442
  8. Desmond DW. The neuropsychology of vascular cognitive impairment. J Neurol Sci. 2004. PMID 15537510
  9. Duering M, et al. Neuroimaging standards for research into small vessel disease—advances since 2013. Lancet Neurol. 2023. PMID 37236211
  10. Wardlaw JM, et al. Neuroimaging standards for research into small vessel disease. Lancet Neurol. 2013. PMID 23867200
  11. Razek AAKA, et al. Imaging of vascular cognitive impairment. Clin Imaging. 2021. PMID 33434866
  12. Swartz RH, et al. Canadian Stroke Best Practice Recommendations: Vascular cognitive impairment, 7th edition update, 2024. Alzheimers Dement. 2025. PMID 39822128
  13. Pendlebury ST, Rothwell PM. Prevalence, incidence, and factors associated with pre- and post-stroke dementia. Lancet Neurol. 2009. PMID 19782001
  14. Barbay M, et al. Prevalence of post-stroke neurocognitive disorders. Dement Geriatr Cogn Disord. 2018. PMID 30504699
  15. Ghafar MZAA, et al. Cognitive screening instruments to identify vascular cognitive impairment. Int J Geriatr Psychiatry. 2019. PMID 31050033
  16. Freitas S, et al. Montreal Cognitive Assessment: validation study for vascular dementia. J Int Neuropsychol Soc. 2012. PMID 22676901
  17. Hachinski VC, et al. Cerebral blood flow in dementia. Arch Neurol. 1975;32:632-7. PMID 1164215
  18. Chui HC, et al. Criteria for the diagnosis of ischemic vascular dementia (ADDTC). Neurology. 1992;42:473-80. PMID 1549205
  19. Román GC, et al. Vascular dementia: diagnostic criteria for research studies (NINDS-AIREN). Neurology. 1993;43:250-60. PMID 8094895
  20. Hachinski V, et al. NINDS-Canadian Stroke Network vascular cognitive impairment harmonization standards. Stroke. 2006;37:2220-41. PMID 16917086
  21. Chui HC, et al. Clinical criteria for the diagnosis of vascular dementia: a multicenter study of comparability and interrater reliability. Arch Neurol. 2000;57:191-6. PMID 10681076
  22. Gold G, et al. Sensitivity and specificity of newly proposed clinical criteria for possible vascular dementia. Neurology. 1997;49:690-4. PMID 9305324
  23. Gold G, et al. Clinicopathological validation study of four sets of clinical criteria for vascular dementia. Am J Psychiatry. 2002;159:82-7. PMID 11772694
  24. Bacchetta JP, et al. Validation of clinical criteria for possible vascular dementia in the oldest-old. Neurobiol Aging. 2007;28:579-85. PMID 16580095
  25. Erkinjuntti T, et al. The effect of different diagnostic criteria on the prevalence of dementia. N Engl J Med. 1997;337:1667-74. PMID 9385127
  26. Pohjasvaara T, et al. Comparison of different clinical criteria (DSM-III, ADDTC, ICD-10, NINDS-AIREN, DSM-IV) for the diagnosis of vascular dementia. Stroke. 2000;31:2952-7. PMID 11108755
  27. Lopez OL, et al. Reliability of NINDS-AIREN clinical criteria for the diagnosis of vascular dementia. Neurology. 1994;44:1240-5. PMID 8035923
  28. Skrobot OA, et al. The Vascular Impairment of Cognition Classification Consensus Study. Alzheimers Dement. 2017;13:624-633. PMID 27960092
  29. Sachdev PS, et al. The Vascular Behavioral and Cognitive Disorders criteria for vascular cognitive disorders: a validation study. Eur J Neurol. 2019;26:1161-1167. PMID 30927497
  30. Folloso MC, et al. The AHA/ASA and DSM-V diagnostic criteria for vascular cognitive impairment identify cases with predominant vascular pathology. Int J Stroke. 2024;19:925-934. PMID 38651759
  31. Knopman DS, et al. Vascular dementia in a population-based autopsy study. Arch Neurol. 2003;60:569-75. PMID 12707071
  32. Cerullo E, et al. Interrater agreement in dementia diagnosis: a systematic review and meta-analysis. Int J Geriatr Psychiatry. 2021;36:1127-1147. PMID 33942363
  33. Erkinjuntti T, et al. Research criteria for subcortical vascular dementia in clinical trials. J Neural Transm Suppl. 2000;59:23-30. PMID 10961414
  34. Erkinjuntti T. Subcortical vascular dementia. Cerebrovasc Dis. 2002;13 Suppl 2:58-60. PMID 11901245
  35. Hsiung GY, et al. Outcomes of cognitively impaired not demented at 2 years in the Canadian Cohort Study of Cognitive Impairment and Related Dementias. Dement Geriatr Cogn Disord. 2006;22:413-20. PMID 16966831
  36. Markus HS, et al. Framework for Clinical Trials in Cerebral Small Vessel Disease (FINESSE). JAMA Neurol. 2022;79:1187-1198. PMID 35969390