Clinical and radiographic subtypes of vascular cognitive impairment in a clinic-based cohort study

Journal of the Neurological Sciences 240 (2006) 7 – 14 www.elsevier.com/locate/jns

Clinical and radiographic subtypes of vascular cognitive impairment in a clinic-based cohort studyB Kenneth Rockwood a,*, Sandra E. Black b, Xiaowei Song a, David B. Hogan c, Serge Gauthier d, Chris MacKnight a, Robert Vandorpe a, Antonio Guzman e, Patrick Montgomery f, Andrew Kertesz g, Remi W. Bouchard h, Howard Feldman i a

Dalhousie University, Canada University of Toronto, Canada c University of Calgary, Canada d McGill University, Canada e University of Ottawa, Canada f University of Manitoba, Canada g University of Western Ontario, Canada h Universite´ Laval, Canada i University of British Columbia, Canada b

Received 25 March 2005; received in revised form 15 July 2005; accepted 22 August 2005 Available online 5 October 2005

Abstract Background and purpose: There is a need for empirical studies to define criteria for vascular cognitive impairment (VCI) subtypes. In this paper, we report the predictive validity of a subtype classification scheme based on clinical and radiographic features. Methods: Nine Canadian memory clinics participated in the Consortium to Investigate Vascular Impairment of Cognition. This cohort consisted of 1347 patients, of whom 324 had VCI, and was followed for up to 30 months. Results: Clinical and neuroimaging features defined three subtypes: vascular cognitive impairment, no dementia, (n = 97), vascular dementia (n = 101) and mixed neurodegenerative/vascular dementia (n = 126). Any ischemic lesion on neuroimaging increased the odds (odds ratio = 9.31; 95% confidence interval 6.46, 13.39) of a VCI diagnosis. No VCI subtype, however, was associated with a specific neuroimaging abnormality. Compared to those with no cognitive impairment, patients with each VCI subtype had higher rates of death and institutionalization (hazard ratio for combined adverse events = 6.08, p < 0.001). Conclusions: Both clinical features and radiographic features help establish a diagnosis of VCI. The outcomes of VCI subtypes, however, are more strongly associated with clinical features than with radiographic ones. D 2005 Elsevier B.V. All rights reserved. Keywords: Vascular cognitive impairment; Neuroimaging; Cerebrovascular disease; Index variables; Validation; Subtypes

i

Kenneth Rockwood was the Principal Investigator of the CIVIC study and wrote the first draft of the paper. Xiaowei Song conducted the analyses and drafted parts of Methods; Robert Vandorpe supervised the neuroimaging and aided in analysis. Chris MacKnight, Serge Gauthier, Antonio Guzman, Patrick Montgomery, Sandra Black, David B Hogan, Andrew Kertesz, Remi Bouchard and Howard Feldman each led the study at their centres. Each read and approved the final version of the manuscript and contributed to interim drafts. In addition, Howard Feldman is the principal investigator of A Collaborative Cohort Of Related Dementias study, conducted in parallel, which gathered and verified initial data from half the sites. * Corresponding author. Centre for Health Care of the Elderly, 5955 Veterans’ Memorial Lane, Suite 1421, Halifax, Nova Scotia, Canada B3H 2E1. Tel.: +1 902 473 8687; fax: +1 902 473 1050. E-mail address: [email protected] (K. Rockwood). 0022-510X/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.jns.2005.08.010

8

K. Rockwood et al. / Journal of the Neurological Sciences 240 (2006) 7 – 14

1. Introduction From ‘‘arteriosclerotic dementia’’ to ‘‘multi-infarct dementia’’ to ‘‘vascular dementia’’ to ‘‘vascular cognitive impairment,’’ to most recently, ‘‘vascular cognitive disorder’’ [1,2], the diagnosis of cognitive and functional impairment seen with cerebrovascular disease continues to evolve. In contrast to earlier consensus-based approaches [3,4], future criteria will be more evidence based [2,5,6]. One candidate starting point for new criteria will be ‘‘vascular cognitive impairment’’ (VCI) [7 –10]. Given the broad conceptualization of VCI, subtypes will need to be identified if the classification of cognitive impairment in relation to cerebrovascular disease is to be clinically useful [7,10 –13]. How to define subtypes is not clear; possible characterizations include risk factors, mechanisms, pathological features, radiographic and laboratory characteristics, clinical features, and/or response to treatment. Three proposed subtypes are patients with vascular dementia (VaD), those with vascular dementia in whom a neurodegenerative dementia also is recognized (i.e., ‘‘mixed’’ neurodegenerative/vascular dementia) and those whose cognitive impairment does not meet dementia criteria (i.e., ‘‘vascular cognitive impairment, no dementia – VCI-ND’’) [14]. Defined clinically, these groups have different outcomes [15], but further diagnostic refinements that emphasize neuroimaging have been proposed [11,12]. To help build an evidence base for clinical and radiographic factors that might aid in subtyping VCI, we report data from the Consortium to Investigate Vascular Impairment of Cognition (CIVIC) study [16]. The CIVIC study’s general aim is to empirically test the various sets of consensus-based criteria [9] by evaluating how VCI is diagnosed. It is important to recognize that this study observed how VCI was diagnosed, rather than prescribed how it should be diagnosed. In short, rather than develop yet another set of consensus criteria, we recognized that dementia specialists diagnosed VCI in practice. Thus, the CIVIC study investigated which of the features from consensus-based criteria practicing dementia specialists actually used. Here, our objective is to test the predictive validity of VCI subtypes. The rationale for focusing on predictive validity is that while the historical referent criterion (the so-called ‘‘gold standard’’ of neuropathology) has lost some of its lustre [12,17,18], the prediction of relevant, nonarbitrary outcomes still enables clinicians to understand the criterion validity of any proposed classification scheme.

2. Methods Following the methods and terminology of Streiner and Norman [19], we evaluated the validity of subtypes by first understanding whether patients looked recognizably different from each other (construct validity) and then contrasted

their outcomes (predictive/criterion validity). We compared characteristics across VCI subtypes to patients with no cognitive impairment (n = 151; note that these people had presented with memory complaints and were not normal volunteers) to those with vascular cognitive impairment, no dementia (VCI-ND, n = 253), that was not of a vascular cause (chiefly, these patients had mild cognitive impairment) or to those with probable AD, diagnosed using standard criteria that excluded mixed AD/VaD [20]. To evaluate predictive validity, we compared rates of death and institutionalization. As detailed elsewhere [16], CIVIC enrolled 1347 patients from 9 Canadian memory clinics. In general, dementia and AD were diagnosed using standard criteria [20,21]. VCI was diagnosed in 324 people (24%). We built on usual care, to which was added a clinical report form. The report form incorporated every unique item from the standard criteria [20,21], the Hachinski Ischemia score [22] and the National Institute of Neurological Disorders and Stroke/Association Internationale pour la Recherche et l’Enseignement en Neurosciences [4], Alzheimer’s Disease Treatment Centers of California [3] and International Classification of Disease, 10th edition [23] criteria. In this way, we were able both to see how many patients met each set of criteria (as reported earlier [16]) and to evaluate which items were most often recorded when a VCI diagnosis was made. In a separate report, we will evaluate each individual criterion; here, we take the necessary first step of presenting data on the major subtypes that have been proposed for clinical and imaging evaluation. In addition to recording data on all VaD criteria, we completed the Disability Assessment for Dementia [24], the Mini-Mental State Examination [25,26], the Functional Assessment Staging Tool [27], the Functional Rating Scale [28] and the Cumulative Illness Rating Scale [29]. In all scores save the Mini-Mental State Examination and Disability Assessment for Dementia, a higher score means worse performance. The CIVIC protocol paralleled those of the Canadian Study of Health and Aging [30] and A Collaborative Cohort of Related Dementias [31]. The checklist and initial clinical classification have proved to be reliable [32]. Neuropsychological testing and neuroimaging were obtained at the discretion of the examining physician. As we built on usual care, we had no standard protocol for the actual imaging acquisition but employed a guide for the interpretation of CT and MRI (available upon request) for scans that were recorded during initial assessments or in the previous 3 months. The interpretation guidelines were based on those used in the Stroke Data Bank [33] and the Dutch TIA Study [34,35] Here, we report CT data only (n = 779, of whom 691 had either no cognitive impairment, non-vascular cognitive impairment without dementia, VCI, or AD; total n = 1191). The CIVIC protocol required regular follow-up over 30 months, either in clinic (where the original assessments

K. Rockwood et al. / Journal of the Neurological Sciences 240 (2006) 7 – 14

9

Table 1 Baseline characteristics of patients with different VCI subtypes, AD, non-vascular CIND, and NCI No cognitive impairment

Age mean (S.D.) Female n (%) Education mean (S.D.) Hypertension n (%) Diabetes n (%) Dyslipidemis n (%) Mini-Mental Score Examination mean (S.D.) Functional Rating Scale mean (S.D.) Disability Assessment for Dementia mean (S.D.)

Non-vascular cognitive impairment, no dementia

Vascular cognitive impairment, no dementia

Vascular dementia

Alzheimer’s disease/vascular dementia

Alzheimer’s disease

n = 151

n = 253

n = 97

n = 101

n = 126

n = 463

62.9 91 12.6 49 15 33 28.6

68.5 140 12.1 85 29 44 26.1

72.8 48 10.6 54 28 27 25.9

75.4 36 11.2 57 19 21 21.5

78.0 63 11.0 68 21 19 19.0

75.6 298 11.0 143 35 59 19.1

(12.8) (60.3) (3.4) (32.5) (9.9) (21.9) (1.9)

(12.0) (55.3) (4.4) (33.6) (11.5) (17.4) (3.8)

(9.9) (49.5) (3.6) (55.7) (28.9) (27.8) (3.2)

(8.1) (35.6) (3.7) (56.4) (18.8) (20.8) (5.6)

(6.7) (50.0) (4.0) (54.0) (16.7) (15.1) (6.1)

(7.9) (64.4) (3.8) (30.9) (7.6) (12.7) (6.3)

F, v 2

p

57.4 34.7 6.3 55.8 43.2 17.6 127.1

<0.001 <0.001 <0.001 <0.001 <0.001 0.003 <0.001

10.6 (2.6)

14.6 (4.6)

14.8 (4.2)

23.4 (6.4)

24.7 (6.2)

23.3 (6.9)

193.8

<0.001

37.4 (4.7)

33.2 (8.0)

32.0 (7.3)

24.8 (10.2)

24.7 (9.3)

26.5 (9.5)

37.1

<0.001

were repeated) or by a telephone interview (which included the informant-based Disability Assessment for Dementia and the Informant Questionnaire on Cognitive Decline in the Elderly [36].) The Canadian Study of Health and Aging decedent interview [37] was used to assess pre-morbid progression of cognitive and functional impairment in patients who had died. In the analysis of construct validity, we first compared demographic and clinical features (including those from the history) between VCI and other diagnoses and within VCI subtypes. Comparisons were made between individuals with NCI, AD, non-vascular cognitive impairment without dementia and the VCI subtypes of VCI-ND, VaD and mixed AD/VaD. To evaluate risks associated with VCI, the comparator group was no cognitive impairment, as the logic is to define a disease entity. We compared VCI subtypes with AD, as the logic is to define a new entity amongst people with cognitive impairment. Baseline intra-group variation was investigated using v 2 and ANOVA.

In the evaluation of predictive validity, we used the logrank test to assess differences in time-dependent outcomes between diagnostic groups. For multivariable modeling of time-dependent outcomes, the assumption of proportional hazards was tested and, if verified, was used to evaluate differences in survival times to death and time to institutionalization, portrayed with Kaplan –Meier curves. In the evaluation of both construct and criterion validity, we were obliged to consider a large number of variables that might define subtypes. Recognizing the power required to evaluate so many factors, dimensionality reduction was achieved in several ways. In addition to multivariable modeling, we combined 20 vascular risk factors (e.g., hypertension, diabetes, dyslipidemia) in a vascular risk factor index variable and 17 clinical features in a vascular clinical profile index variable, as described elsewhere [38]. The clinical index included acute onset, fluctuating course, periods of prolonged plateaus, memory impairment not the main complaint, focal symptoms, frequent falls, an early gait abnormality, early urinary incontinence, pseudobulbar

Table 2 Vascular profiles of patients with different VCI subtypes, AD, non-vascular CIND, and NCI Number of present (%)

Acute onset Stepwise progression Fluctuating course Nocturnal confusion Gait abnormality Urinary frequency History of falls Personality changes Mood changes Emotional incontinence Somatic complaints Psychomotor retardation

No cognitive impairment

Non-vascular cognitive impairment, no dementia

Vascular cognitive impairment, no dementia

Vascular dementia

Alzheimer’s disease/vascular dementia

Alzheimer’s disease

n = 151

n = 253

n = 97

n = 101

n = 126

n = 463

1 0 0 0 1 1 3 5 8 0 6 2

3 0 4 3 12 8 19 28 40 2 4 3

13 3 8 3 14 12 18 19 26 5 2 2

18 12 11 12 28 22 30 31 29 10 3 12

5 6 5 20 14 16 16 24 25 6 3 7

4 1 5 17 15 19 21 71 66 7 7 8

(0.7)

(0.7) (0.7) (2.0) (3.3) (5.3) (4.0) (1. 3)

(1.2) (1.6) (1.2) (4.8) (3.2) (7.5) (11.1) (15.8) (0.8) (1.6) (1.2)

(13.4) (3.1) (8.2) (3.1) (14.4) (12.4) (18.6) (19.6) (26.8) (5.2) (2.1) (2.1)

(17.8) (11.9) (10.9) (11.9) (27.7) (21.8) (29.7) (30.7) (28.7) (9.9) (3.0) (11.9)

(4.0) (4.8) (4.0) (15.9) (11.1) (12.7) (12.7) (19.0) (19.8) (4.8) (2.4) (5.6)

(0.9) (0.0) (1.1) (3.7) (3.2) (4.1) (4.5) (15.3) (14.3) (1.5) (1.5) (1.7)

v2

p

63.1 40.6 26.1 30.1 26.2 14.4 21.0 4.4 11.9 9.1 36.4 11.6

<0.001 <0.001 <0.001 <0.001 <0.001 0.013 0.001 0.498 0.037 0.106 <0.001 0.041

10

K. Rockwood et al. / Journal of the Neurological Sciences 240 (2006) 7 – 14

Table 3 Radiographic neuroimaging lesion profiles of patients with different VCI subtypes, AD, non-vascular CIND, and NCI

Cortical Subcortical White matter Cortical and white matter Cortical and subcortical Subcortical and white matter Other intracranial lesions All of these types None of these types

No cognitive impairment

Non-vascular cognitive impairment, no dementia

n = 37

n = 133

n = 77

2 3 8 0 0 1 4 0 19

0 3 36 0 0 0 12 0 82

18 (23.4) 8 (10.4) 22 (28.6) 3 (3.9) 1 (1.3) 9 (11.7) 4 (5.2) 0 12 (15.6)

(5.4) (8.1) (21.6)

(2.7) (10.8) (51.4)

(2.3) (27.1)

(9.0) (61.7)

palsy, low mood and patchy cognitive deficits. The index variables are each scaled as unweighted proportions (e.g., in the 17-item clinical index, a patient with 4 items would have an index score of 4/17 = 0.24). As reported elsewhere, the shape and scale parameters of their distributions suggest that the index variables operate validly as state variables; their observed ranges were 0 –0.58 and 0– 0.65 [38]. Each index was evaluated in multivariable models. Final models to define subtypes were based on both statistical and clinical criteria. For example, factors not available at baseline, (e.g., progression of impairment) or not routinely available (e.g., MRI) or that while traditional criteria were little used by clinicians (e.g., ‘‘preservation of personality’’), were not entered in the final model. In addition, as vascular risk factors are recognized as risks for AD [39], the risk factor index variable was not included in the logistic regression model for assessments of subtypes, where AD patients formed the comparator group. By contrast, given their status as risk factors for impaired cognition [40], the vascular risk factor index variable was included in the model for VCI, in which NCI patients formed the comparator group.

Vascular cognitive impairment, no dementia

Vascular dementia

Alzheimer’s disease/vascular dementia

Alzheimer’s disease

Total

n = 87

n = 99

n = 258

n = 691

17 19 23 7 3 8 1 0 9

15 13 32 2 5 13 6 0 12

5 11 86 0 0 1 17 1 138

57 57 207 12 9 32 44 1 272

(19.5) (21.8) (26.4) (8.1) (3.5) (9.2) (1.2) (10.3)

(15.2) (13.1) (32.3) (2.0) (5.1) (13.1) (6.1) (12.1)

(1.9) (4.3) (33.3)

(0.4) (6.6) (53.5)

(8.2) (8.2) (30.0) (1.7) (1.3) (4.6) (6.4) (0.1) (39.4)

similar to each other, and to patients with AD, than to patients with no cognitive impairment (Table 1). More VCI patients were men and were generally more likely to have vascular risk factors. VCI patients differed from patients with AD, and between each other, by clinical subtypes. Of the clinical features, acute onset, gait abnormalities, falls and urinary abnormalities were most associated with a VCI diagnosis. These remain associated with VCI in models adjusted for age, sex and degree of severity (Table 2).

(A) Survival function estimate

Number of present (%)

1.00

NCI

0.95 AD

0.90 0.85 NCI (n=151) Vasc. CIND (n=151) VaD (n=101) Ad/VaD (n=101) Ad (n=463)

0.80 0.75 0

10

20

30

Time to death (months)

3. Results

Table 4 Factors associated with a diagnosis of vascular cognitive impairment compared with no cognitive impairment in a multivariable logistic regression analysis

Age Sex Years of education MMSE Risk index Clinical index Lesion Constant

Odds ratio (95% CI)

Wald

Sig

1.02 (0.99 – 1.04) 0.53 (0.37 – 0.76) 0.99 (0.94 – 1.03) 1.05 (1.02 – 1.08) 1.07 (1.05 – 1.09) 1.06 (1.04 – 1.08) 9.20 (6.39 – 13.23) 0.007

2.507 11.937 0.308 9.386 44.919 51.568 142.984 29.561

0.113 0.001 0.579 0.002 <0.001 <0.001 <0.001 <0.001

Survival function estimate

(B) In evaluating whether subtypes are recognizably distinct, we first note that the 324 patients with VCI were more

1.00 NCI

0.95 0.90 0.85

AD

0.80 0.75 0

10

20

30

Time to institutionalization (months) Fig. 1. Survival curves for mortality (Panel A) and institutionalization (Panel B) by clinico-radiographic subtype.

K. Rockwood et al. / Journal of the Neurological Sciences 240 (2006) 7 – 14

Neuroimaging profiles show modest but statistically significant differences across diagnostic categories (Table 3), although within VCI subtypes, the distributions of lesions were similar. In models of factors associated with a diagnosis of VCI, several features were important, including, most obviously, measures of the degree of cognitive and functional impairment (Table 4). For example, every 1-point decrement from 30 in the MMSE increased the odds of a VCI-ND diagnosis by 1.05 (95% confidence interval 1.02 –1.08). Every 1-point change in the vascular risk index profile showed a similarly increased likelihood (1.07; 1.05– 1.09) as did a 1-point increase in the clinical index variable (1.06; 1.04– 1.08). An ischemic lesion, particularly stroke detected by neuroimaging, increased the risk by 5.21 (2.96 – 7.45) when other factors are controlled. Similar profiles of vascular risks, vascular clinical features and ischemic neuroimaging lesions increased the risk for the subtypes of vascular dementia and mixed dementia. Cortical or subcortical stroke, rather than white matter changes, increased the odds of a VCI diagnosis. Both clinical and imaging subtypes were associated with adverse outcomes (Figs. 1 and 2). All cognitively impaired

Survival function estimate

(A)

1.00 0.95 0.90 0.85 No leision (n=33) No white matter leision (n=110) Any white matter leision (n=120)

0.80 0.75 0

10

20

30 3

Time to death (months)

(B) Survival function estimate

1.00 0.95 0.90 0.85 0.80 0.75 0

10 20 Time to institutionalization (months)

30

Fig. 2. Kaplan – Meier survival estimates of the time to death (Panel A) and institutionalization (Panel B) amongst patients with vascular cognitive impairment, by imaging subtype (here grouped as any lesion with white matter involvement, any lesion without white matter involvement, and no imaged lesion).

11

patients had higher rates (cf. no cognitive impairment) of death and institutionalization (log-rank test = 33.3, P < 0.001). People with the subtypes VaD and mixed dementia had higher 30-month rates of adverse outcomes. By contrast, even compared to people with no cognitive impairment and no imaged lesions, no single type of neuroimaging lesion conveyed a worse prognosis. For example, of the 207 patients with white matter lesions, 21 were institutionalized and 17 died (total adverse event rate 18%), compared with 30 (24%) adverse outcomes (including 12 deaths) amongst the 123 with ischemic lesions without white matter involvement and 6 (18%) adverse events amongst those with VCI but no imaged lesion (logrank test = 0.09, p = 0.95), A similar result is obtained for institutionalization. (Note that the survival estimates cross over the full time interval, so that the assumption of proportionality of hazards is violated and, hence, the presentation only of unadjusted estimates.)

4. Discussion In a multi-centre, Canadian memory clinic-based study, we found that both clinical features and radiographic features helped to establish a diagnosis of VCI, selecting patients who were recognizably different from those with no cognitive impairment and with AD. Both clinical and radiographic features also allowed VCI subtypes to be defined, but the outcomes of patients with differing VCI subtypes were more strongly associated with clinical features than with radiographic ones. Given that VCI is common [15], costly [41] and perhaps susceptible to treatment or even prevention [39,42], our finding that it can be readily classified and pragmatically sub-classified can help to establish a framework for future studies. Such sub-classification is needed, given that VCI is broadly constituted and that sub-groups of clinically identifiable patients might usefully guide diagnosis and management. Our data must be interpreted with caution. Estimates derived from memory clinics will differ from estimates derived from stroke clinics [6]. This does not mean that they will be unimportant, however, especially if VCI without clinically evident stroke is more common than VCI with stroke [11,15]. The most important limitation, however, is the absence of a study-specific neuroimaging protocol for all patients. In particular, of the 1191 patients whom we considered as VCI, no cognitive impairment, non-vascular cognitive impairment without dementia or AD, only 691 (58%) had relevant neuroimaging within 3 months of the initial clinic visit. Selection for imaging was biased towards VCI (81% of whom had CTs) compared with no cognitive impairment (25%), non-vascular CIND (53%) or AD (56%). On the other hand, the numbers of scans that we considered are substantial enough that strong signals should be readily detected. Moreover, the impact of imaging recruitment bias

12

K. Rockwood et al. / Journal of the Neurological Sciences 240 (2006) 7 – 14

would be to over-estimate the number of people with lesions, as well as their impact. Thus, our estimate that, even amongst imaged patients, an important minority with clinically diagnosed VCI did not have CT neuroimaging evidence of an ischemic lesion is a conservative one. Importantly, too, in the unadjusted analyses, this profile conferred a poor prognosis. How to interpret a ‘‘clinical VCI/no CT-imaged lesion’’ subclass is unclear. On the one hand, it might simply represent falsely negative neuroimaging, inasmuch as routine MRI might have shown important ischemic lesions not evident on CT [43]. On the other hand, while CT/MRI helps to validate the clinical diagnosis, it is clear that clinically important ischemic lesions can exist which are not well correlated with available neuroimaging [44] (i.e., that there are problems of specificity—also evident in the observation that MRI-defined white matter changes are common in all types of VaD) [45]. Additionally, there are problems of sensitivity—the radiological portion of VaD diagnostic criteria now exclude bona fide cases of VaD [46]. Similarly, although neuropsychological testing occurred at each site, it was ordered variably and included variable tests, so that standard neuropsychological data cannot be reported for all VCI patients. Against this background, it should not be surprising that both clinical and neuroimaging features are important to a VCI diagnosis. Consider too that even in post-stroke cohorts, both imaging (CT) and clinical features are important in predicting dementia [47]. Moreover, although there is clearly room for methodological improvement in many of the studies that have not discerned a specific relationship between lesions and cognitive deficits [48], even longitudinal studies have not always been able to disentangle the associations between the neuroimaging and the cognitive deficit [49]. In this context, our pragmatic clinical focus suggests that while the neuroimaged lesion conveys information helpful to a VCI diagnosis, additional data are needed to understand prognosis. While it might be that more sensitive modalities, including indices of cerebral metabolism which might have a better yield, some of the claims made for advanced imaging techniques are faltering. For example, the distinction between periventricular white matter hyperintensities and deep ones has been found to be less helpful than considering total WMH volume relationships [50]. Other recent studies have undermined the extent of the specificity of localized MRI-detectable lesions in the presentation of dementia. For example, regardless of where in the brain white matter hyperintensities are located, they have been found to be associated with frontal hypometabolism and executive dysfunction [51]. In addition, the neuroimaging component of the National Institute of Neurological Disorders and Stroke/Association Internationale pour la Recherche et l’Enseignement en Neurosciences criteria does not distinguish between older patients with and without post-stroke dementia [52]. Thus,

while better imaging might have transformative effects, it is also important to assess what might be done in the meantime, especially outside large academic centres, and in non-Western countries, where CT might be the best that is routinely available [53]. How to understand neuroimaging illustrates a more general problem that has proved to be difficult in developing VCI criteria, in that single factors can operate as both exposures and outcomes. For example, white matter changes are common in all dementia types examined— and not rare in older people without cognitive impairment [54]. White matter changes are importantly related to disease exposures (e.g., hypertension) and in this sense operate as an outcome. In addition, however, they are a risk factor for adverse outcomes (e.g., impaired cognition) and thus also operate as an exposure [55 – 58]. The need not to include vascular risk factors as criteria for VCI is well recognized, but these data call attention to how circularity might be avoided with respect to neuroimaging. This must be addressed forthrightly, as new neuroimaging techniques are being developed against a background of questioning the role of neuropathology [18]. The frequent disagreement between differing neuropathological criteria [59], the number of people in whom a satisfactory neuropathological diagnosis cannot be established [60] and the large contribution by vascular lesions to neurodegenerative disease expression [61 – 63] argue for caution in the role accorded post-mortem examination of the brain. On the other hand, given the ambivalent status of some lesions as exposures and outcomes and the relationship between an adverse vascular profile and adverse outcomes in the absence of neuroimaged lesions, it would seem wrong to toss out neuropathology as a ‘‘gold standard’’ only to embrace neuroimaging. The CIVIC experience suggests that neuroimaging lesions should be seen as supportive –even highly supportive (for example, necessary for a ‘‘probable’’ diagnosis) – of VCI, but not essential. Recently, a proposal has been made to, in effect, set aside most of the evidence about VCI in favour of the newly introduced concept of ‘‘vascular cognitive disorder’’ (VCD) [2]. The essential difference between VCI and VCD is in the approach to the diagnosis of dementia and in the rejection of a gradient of impairment in favour of ‘‘bona fide irrefutable cases of the disease.’’ How the VCI/VCD debate will evolve remains unclear. For now, the existence of the idea that a new approach might be preferable is not a persuasive argument to ignore empirical studies. Nevertheless, VCI/ VCD/VaD is clearly an area in which additional debate (and not just additional data) can help illuminate the tricky but tractable question of how cerebrovascular disease affects cognition. If VCI is to be understood as a broadly defined construct, sub-grouping is necessary. In this memory-clinic based study, we found that both clinical and radiographic features were important in defining VCI and in identifying subtypes. Once clinical subtypes had been established, however,

K. Rockwood et al. / Journal of the Neurological Sciences 240 (2006) 7 – 14

meaningful differences in death and institutionalization were more strongly associated with clinical characteristics than with neuroimaging features.

5. Competing interest statement This is an original work that is not submitted elsewhere. Each of the coauthors has made a substantial contribution to the paper, as outlined. The study was sponsored by the Canadian Institutes of Health Research and in part by its predecessor, the Medical Research Council of Canada, then through a partnership program with the company once known as Hoechst Marion Roussel. None of the sponsors has had any role in the initiation of the work, the data collection, or the analysis of the data, which are located with and analysed by my group, allowing me to take public responsibility for the entire work.

Acknowledgements The CIVIC study was funded by grants from the Medical Research Council of Canada through the PMAC/MRC program, with support from Hoechst Marion Roussel Canada, and by the Alzheimer Society of Canada. Additional funding for these analyses came from the Canadian Institutes of Health Research (CIHR) grant number MOP 62823. Kenneth Rockwood and Chris MacKnight receive support from the CIHR through Investigator and New Investigator awards, respectively. Kenneth Rockwood is also supported by the Dalhousie Medical Research Foundation as Kathryn Allen Weldon Professor of Alzheimer Research. David Hogan receives career support as the Brenda Strafford Foundation Chair in Geriatric Medicine at the University of Calgary.

References [1] Sachdev P. Vascular cognitive disorder. Int J Geriatr Psychiatry 1999;14:402 – 3. [2] Roma´n GC, Sachdev P, Royall DR, Bullock RA, Orgogozo JM, Lopez-Pousa S, et al. Vascular cognitive disorder: a new diagnostic category updating vascular cognitive impairment and vascular dementia. J Neurol Sci 2004;226:81 – 7. [3] Chui HC, Victoroff JI, Margolin D, Jagust W, Shankle R, Katzman R. Criteria for the diagnosis of ischemic vascular dementia proposed by the state of California Alzheimer’s Disease Diagnostic and Treatment Centers. Neurology 1992;42:473 – 80. [4] Roma´n GC, Tatemichi TK, Erkinjuntti T, Cummings JL, Masdeu JC, Garcia JH, et al. Vascular dementia: diagnostic criteria for research studies. Report of the NINDS-AIREN International Workshop. Neurology 1993;43:650 – 60. [5] Chui HC, Mack W, Jackson JE, Mungas D, Reed BR, Tinklenberg J, et al. Clinical criteria for the diagnosis of vascular dementia: a multicenter study of comparability and interrater reliability. Arch Neurol 2000;57:191 – 6.

13

[6] Bowler JV. Criteria for vascular dementia: replacing dogma with data. Arch Neurol 2000;57:170 – 1. [7] Erkinjuntti T, Hachinski VC. Rethinking vascular dementia. Cerebrovasc Dis 1993;3:3 – 23. [8] Hachinski V. Vascular dementia: a radical redefinition. Dementia 1994;5:130 – 2. [9] Rockwood K, Parhad I, Hachinski V, Erkinjuntti T, Rewcastle B, Kertesz A, et al. Diagnosis of vascular dementia: consortium of Canadian Centres for Clinical Cognitive Research consensus statement. Can J Neurol Sci 1994;21:358 – 64. [10] Bowler JV, Steenhuis R, Hachinski V. Conceptual background to vascular cognitive impairment. Alzheimer Dis Assoc Disord 1999; 13(Suppl. 3):S30 – 7. [11] Erkinjuntti T, Inzitari D, Pantoni L, Wallin A, Scheltens P, Rockwood K, et al. Research criteria for subcortical vascular dementia in clinical trials. J Neural Transm Suppl 2000;59:23 – 30. [12] O’Brien JT, Erkinjuntti T, Reisberg B, Roman G, Sawada T, Pantoni L, et al. Vascular cognitive impairment. Lancet Neurol 2003;2:89 – 98. [13] Starkstein SE, Almeida OP. Understanding cognitive impairment and dementia: stroke studies. Curr Opin Psychiatr 2003;16:615 – 20. [14] Rockwood K, Howard K, MacKnight C, Darvesh S. Spectrum of disease in vascular cognitive impairment. Neuroepidemiology 1999; 18:248 – 54. [15] Rockwood K, Wentzel C, Hachinski V, Hogan DB, MacKnight C, McDowell I. Prevalence and outcomes of vascular cognitive impairment. Vascular Cognitive Impairment Investigators of the Canadian Study of Health and Aging. Neurology 2000;54:447 – 51. [16] Rockwood K, Davis H, MacKnight C, Vandorpe R, Gauthier S, Guzman A, et al. The Consortium to Investigate Vascular Impairment of Cognition: background, methods and impact of diagnostic criteria on the study of vascular cognitive impairment. Can J Neurol Sci 2003; 30:237 – 43. [17] Bowler JV, Munoz DG, Merskey H, Hachinski V. Fallacies in the pathological confirmation of the diagnosis of Alzheimer’s disease. J Neurol Neurosurg Psychiatry 1998;64:18 – 24. [18] Kalaria RN, Kenny RA, Ballard CG, Perry R, Ince P, Polvikoski T. Towards defining the neuropathological substrates of vascular dementia. J Neurol Sci 2004;226:75 – 80. [19] Streiner D, Norman G. Health measurement scales: a practical guide to their intervention and use. 3rd ed. Oxford’ Oxford University Press; 2003. [20] McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer’s disease: report of the NINCDSADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 1984;34:939 – 44. [21] American Psychiatric Association. Diagnostic and statistical manual. 3rd ed., revised. Washington (DC)’ APA; 1987. [22] Hachinski VC, Iliff LD, Zilhka E, Du Boulay GH, McAllister VL, Marshall J, et al. Cerebral blood flow in dementia. Arch Neurol 1975; 32:632 – 7. [23] World Health Organization International Classification of Diseases. The ICD-10 classification of mental and behavioural disorders: clinical descriptors and diagnostic guidelines. 10th edition. Geneva’ WHO; 1992. [24] Ge´linas I, Gauthier L, McIntyre M, Gauthier S. Development of a functional measure for persons with Alzheimer’s disease: the Disability Assessment for Dementia. Am J Occup Ther 1999;53: 471 – 81. [25] Folstein MF, Folstein SE, McHugh PR. The Folstein Mini-Mental State Examination: a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975;12:189 – 98. [26] Grace J, Nadler JD, White DA, Guilmette TJ, Giuliano AJ, Monsch AU, et al. Folstein vs modified Mini-Mental State Examination in geriatric stroke: stability, validity, and screening utility. Arch Neurol 1995;52:477 – 84.

14

K. Rockwood et al. / Journal of the Neurological Sciences 240 (2006) 7 – 14

[27] Reisberg B. Functional Assessment Staging (FAST). Psychopharmacol Bull 1988;24:653 – 9. [28] Feldman H, Schulzer M, Wang S, Tuokko H, Beattie BL. The Functional Rating Scale in Alzheimer’s disease assessment: a longitudinal study. In: Iqbal K, Mortimer JA, Winblad B, Wisniewski HM, editors. Research advances in Alzheimer’s disease and related disorders. Chichester (UK)’ Wiley; 1995. p. 235 – 41. [29] Conwell Y, Forbes NT, Cox C, Caine ED. Validation of a measure of physical illness burden at autopsy: the Cumulative Illness Rating Scale. J Am Geriatr Soc 1993;41:38 – 41. [30] Rockwood K, McDowell I, Wolfson C, editors. Canadian study of health and aging. Technical papersInt Psychogeriatr, vol. 13 (Suppl. 1). 2001. p. 1 – 237. [31] Feldman H, Levy AR, Hsiung GY, Peters KR, Donald A, Black SE, et al. A Canadian Cohort Study of Cognitive Impairment and Related Dementias (ACCORD): study methods and baseline results. Neuroepidemiol 2003;22:265 – 74. [32] Wentzel C, Darvesh S, MacKnight C, Shea C, Rockwood K. Interrater reliability of the diagnosis of vascular cognitive impairment at a memory clinic. Neuroepidemiology 2000;19:186 – 93. [33] Shinar D, Gross CR, Hier DB, Caplan LR, Mohr JP, Price TR, et al. Interobserver reliability in the interpretation of computed tomographic scans of stroke patients. Arch Neurol 1987;44:149 – 55. [34] van Swieten JC, Hijdra A, Koudstaal PJ, van Gijn J. Grading white matter lesions on CT and MRI: a simple scale. J Neurol Neurosurg Psychiatry 1990;53:1080 – 3. [35] van Swieten JC, Kappelle LJ, Algra A, van Latum JC, Koudstaal PJ, van Gijn J. Hypodensity of the cerebral white matter in patients with transient ischemic attack or minor stroke: influence on the rate of subsequent stroke. Dutch TIA Trial Study Group. Ann Neurol 1992; 32:177 – 83. [36] Jorm AF. A short form of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE): development and cross-validation. Psychol Med 1994;24:145 – 53. [37] Stewart M, McDowell I, Hill G, Aylesworth R. Estimating antemortem cognitive status of deceased subjects in a longitudinal study of dementia. Int Psychogeriatr 2001;13(Suppl. 1):99 – 106. [38] Song X, Mitnitski A, Rockwood K. Index variables for studying outcomes in vascular cognitive impairment. Neuroepidemiology 2005;25:196 – 204. [39] Launer LJ. The epidemiologic study of dementia: a life-long quest? Neurobiol Aging 2005;26:335 – 40. [40] Elias MF, Sullivan LM, D’Agostino RB, Elias PK, Beiser A, Au R, et al. Framingham stroke risk profile and lowered cognitive performance. Stroke 2004;35:404 – 9. [41] Rockwood K, Brown M, Merry H, Sketris I, Fisk J. Societal costs of vascular cognitive impairment in older adults. Stroke 2002;33: 1605 – 9. [42] Erkinjuntti T, Roma´n G, Gauthier S. Treatment of vascular dementia—evidence from clinical trials with cholinesterase inhibitors. J Neurol Sci 2004;226:9 – 12. [43] Scheltens P, Korf ES. Contribution of neuroimaging in the diagnosis of Alzheimer’s disease and other dementias. Curr Opin Neurol 2000; 13:391 – 6. [44] Bowler JV. Vascular cognitive impairment. Stroke 2004;35:386 – 8. [45] Cosentino SA, Jefferson AL, Carey M, Price CC, Davis-Garrett K, Swenson R, et al. The clinical diagnosis of vascular dementia: a comparison among four classification systems and a proposal for a new paradigm. Clin Neuropsychol 2004;18:6 – 21.

[46] van Straaten EC, Scheltens P, Barkof F. MRI and CT in the diagnosis of vascular dementia. J Neurol Sci 2004;226:9 – 12. [47] Rasquin SM, Verhey FR, van Oostenbrugge RJ, Lousberg R, Lodder J. Demographic and CT scan features related to cognitive impairment in the first year after stroke. J Neurol Neurosurg Psychiatry 2004; 75:1562 – 7. [48] Ferro JM, Madureira S. Age-related white matter changes and cognitive impairment. J Neurol Sci 2002;203 – 204:221 – 5. [49] Schmidt R, Schmidt H, Kapeller P, Enzinger C, Ropele S, Saurugg R, et al. The natural course of MRI white matter hyperintensities. J Neurol Sci 2002;203 – 204:253 – 7. [50] DeCarli C, Fletcher E, Ramey V, Harvey D, Jagust WJ. Anatomical mapping of white matter hyperintensities (WMH): exploring the relationships between periventricular WMH, deep WMH, and total WMH burden. Stroke 2005;36:50 – 5. [51] Tullberg M, Fletcher E, DeCarli C, Mungas D, Reed BR, Harvey DJ, et al. White matter lesions impair frontal lobe function regardless of their location. Neurology 2004;63:246 – 53. [52] Ballard CG, Burton EJ, Barber R, Stephens S, Kenny RA, Kalaria RN, et al. NINDS AIREN neuroimaging criteria do not distinguish stroke patients with and without dementia. Neurology 2004;63:983 – 8. [53] Chen CP. Transcultural expression of subcortical vascular disease. J Neurol Sci 2004;226:45 – 7. [54] Pugh KG, Lipsitz LA. The microvascular frontal-subcortical syndrome of aging. Neurobiol Aging 2002;23:421 – 31. [55] Liebetrau M, Steen B, Hamann GF, Skoog I. Silent and symptomatic infarcts on cranial computerized tomography in relation to dementia and mortality. A population-based study in 85-year-old subjects. Stroke 2004;35:1816 – 20. [56] Linden T, Skoog I, Fagerberg B, Steen B, Blomstrand C. Cognitive impairment and dementia 20 months after stroke. Neuroepidemiol 2004;23:45 – 52. [57] Lopez OL, Becker JT, Jungreis CA, Rezek D, Estol C, Boller F, et al. Computed tomography – but not magnetic resonance imaging – identified periventricular white-matter lesions predict symptomatic cerebrovascular disease in probable Alzheimer’s disease. Arch Neurol 1995; 52:659 – 64. [58] de Leeuw FE, de Groot JC, Oudkerk M, Witteman JC, Hofman A, van Gijn J, et al. Hypertension and cerebral white matter lesions in a prospective cohort study. Brain 2002;125:765 – 72. [59] Knopman DS, Parisi JE, Salviati A, Floriach-Robert M, Boeve BF, Ivnik RJ, et al. Neuropathology of cognitively normal elderly. J Neuropathol Exp Neurol 2003;62:1087 – 95. [60] Crystal HA, Dickson D, Davies P, Masur D, Grober E, Lipton RB. The relative frequency of ‘‘dementia of unknown etiology’’ increases with age and is nearly 50% in nonagenarians. Arch Neurol 2000; 57:713 – 9. [61] Medical Research Council Cognitive Function and Ageing Study. Pathological correlates of late-onset dementia in a multicentre, community-based population in England and Wales. Neuropathology Group of the Medical Research Council Cognitive Function and Ageing Study (MRC CFAS). Lancet 2001;357:169 – 75. [62] Snowdon DA, Kemper SJ, Mortimer JA, Greiner LH, Wekstein DR, Markesbery WR. Brain infarction and the clinical expression of Alzheimer disease: the Nun Study. JAMA 1997;277:813 – 7. [63] Zekry D, Duyckaerts C, Belmin J, Geoffre C, Moulias R, Hauw JJ. Alzheimer’s disease and brain infarcts in the elderly. Agreement with neuropathology. J Neurol 2002;249:1529 – 34.