Because the risk of dementia increases with age, as life expectancy has improved, dementia has become an important public health problem. Vascular dementia (VaD) is one of the leading causes of dementia after AD itself.1 AD occurs alone or with cerebrovascular disease (AD+CVD, previously called mixed dementia). In autopsy series, the prevalence of VaD varies from 0 to 58% (likely values 8–15%) of dementia cases; the prevalence of AD+CVD may range from 2 to 36% (likely values 12–18%) of dementia cases.2

More up-to-date knowledge about the clinical progression of VaD and AD+CVD, specifically rates of cognitive decline and survival, would be useful for informing families, assessing the effect of interventions, such as pharmaceutical therapies and providing data for public health planning purposes. Numerous studies describe the clinical features and course of AD alone; far fewer are available for VaD and AD+CVD, despite their high prevalence. In their literature review, Chui and Gonthier identified 13 studies published between 1985 and 1998 that compared survival of AD and VaD patients.3 Since then, however, only a few new survival studies have been published, and their findings are contradictory.4^–8 Similarly, few studies have followed VaD or AD+CVD patients longitudinally to assess the course of their cognitive decline, and those studies have also produced conflicting results.3 9^–13 Moreover, none of the most recent of those studies followed cognitive decline in patients who had AD+CVD.10 11 The most likely explanation for the contradictory findings in the literature is the lack of consensus about the diagnostic criteria for both VaD and AD+CVD, due in part to their clinical and pathological heterogeneity and the multiple pathological subtypes. Thus, some studies used the Hachinski Ischaemia Score,14 others the NINDS-AIREN criteria15 or the ICD criteria16 for VaD diagnosis. In 2004, Roman et al proposed the concept of “vascular cognitive disorder” to group impairments and dementia with a vascular component into one entity. It includes “vascular cognitive impairment,” characterised by a specific cognitive profile involving preserved memory with impairments in attentional and executive functioning, VaD, including cases of dementia resulting from cerebral hypoxic ischaemic lesions or haemorrhagic brain lesions, and AD+CVD, characterised by the existence of AD with vascular lesions that contribute to the alteration of cognitive functions.17 18 Moreover, the diagnosis of dementia after a stroke is complicated because dementia is not always a direct consequence of cerebrovascular lesions;19 in some cases, it may be due instead to the cumulative effect of cerebrovascular and degenerative lesions.

Many patients admitted for stroke have pre-existing cognitive decline or undiagnosed dementia, which means that their poststroke dementia may be due to degenerative lesions.20 21 Indeed, one-third of poststroke dementia cases are due to AD.20 22^–24 The risk of erroneous VaD diagnosis is therefore not insubstantial.25 Neuroimaging helps significantly to limit misclassifications,26 27 but although some recent studies have performed neuroimaging for their patient, the practice is not generalised.6 7 Differences in populations under study, population- versus clinic-based, may also explain the conflicting results in the literature. Clinic-based subjects may not be representative of the general population; for example, they appear to be better educated than population-based subjects. On the other hand, clinic-based studies offer more finely characterised diagnoses; in particular, neuroimaging is available more often in clinic-based studies. Moreover, other clinical characteristics, such as vascular risk factors, are more easily measured in clinic-based settings. While clinic-based samples may not be the most appropriate design for estimating the incidence of dementia in the population, they offer the possibility of following well-characterised demented patients and taking their clinical features into account in analyses. Only four recent studies have clinic-based settings:4 8 10 11 they also had limited populations, no systematic MRI and conflicting results.

The aim of this clinical cohort study was to compare demographic characteristics at baseline, survival and cognitive decline in patients with probable and definite VaD, AD+CVD and AD alone, among 970 outpatients who had brain imaging as part of their follow-up at the Lille-Bailleul memory clinic.

METHODS

Subjects

Between 1995 and 2001, 3545 new consecutive patients diagnosed as having VaD, AD+CVD and AD were examined at the Lille-Bailleul memory clinic. Excluded from this study are the 2575 patients with a diagnosis of “possible” VaD (n = 213) or AD (n = 614), patients not examined at least twice during a period of at least 1 year (n = 1602) and patients without a Dementia Rating Scale (DRS) measurement during the year of their first visit (n = 146). The Lille memory clinic is a research and resource clinic; it has a local catchment zone but also serves as a resource and reference centre for the region, which explains that not all the patients examined are followed up by the clinic. In all, 970 patients met the inclusion criteria: 14.5% had diagnoses of “probable” or “definite” VaD, 17.1% of AD+CVD and 68.4% of AD. Patients in the study sample were followed for a mean of 4.7 (SD 2.7) years. The mean age of the population was 73 (8) years and differed significantly from that of the population excluded (74 (9) years, p = 0.0004). Women accounted for 64.9% of this population and did not differ significantly from the excluded population (62.6%, p = 0.22). Excluded patients had AD (76.0%) and VaD (17.4%) more often, and AD+CVD (6.6%) less often than the patients studied here (p<0.0001). The mean MMSE and DRS scores were higher in the study sample than in the excluded population (mean MMSE: 21.0 (6) vs 17.5 (8), p<0.0001; mean DRS: 110.9 (26) vs 97.5 (40), p<0.0001). Study patients had hypertension (36.0% vs 20.8%, p<0.0001) and diabetes (11.1% vs 7.0%, p<0.0001) more often than the excluded group.

Procedures

Diagnostic evaluation

A multidisciplinary staff conducted a standardised examination including neurological, neuropsychological, behavioural, laboratory and imaging assessments for each patient. Whenever possible, a relative or other informant was present at the interview. Subjects with AD had to meet NINCDS-ADRDA (National Institute of Neurologic and Communicative Disorders and Stroke and the AD and Related Disorders Association) criteria for diagnosis of “probable” or “definite” AD28 and subjects with VaD the NINDS-AIREN criteria (Neuroepidemiology branch of the US National Institute of Neurological Disorders and Stroke and the Association Internationale pour la Recherche et l’Enseignement en Neurosciences) for the diagnosis of “probable” VaD.15 Subcortical ischaemic vascular dementia was defined according to more recent criteria and was found in 68.8% of our VaD sample (n = 97).17 Subjects who had AD+CVD had to meet the clinical criteria for possible AD and present clinical or brain imaging evidence of relevant CVD. More specifically, we classified patients as having AD+CVD (or “Mixed dementia”) when:

1. there was a clinical history suggestive of “probable Alzheimer disease;” a stepwise progression is suggestive of cerebrovascular disease but not mandatory since progressive cognitive impairment may occur in extensive white-matter lesions.

2. there was a normal clinical examination or focal neurological signs suggestive of cerebrovascular disease;29

3. there was a neuropsychological profile suggestive of “probable Alzheimer disease” with progressive impairment of episodic memory and two or more other domains of cognition (impairment of encoding, recall and storage of information, with a delayed recall worse than the immediate recall), associated with a pattern of cognitive deficit suggesting vascular lesions with prominent executive dysfunction (slowed information processing, impairment in the ability to shift from one task to another, and deficits in the ability to hold an manipulate information) more severe than expected in “pure” probable Alzheimer disease;

4. MRI scans showed abnormalities seen in Alzheimer disease with medial temporal lobe atrophy, and relevant cerebrovascular disease including large-vessel infarct, or strategic infarct (thalamus or basal ganglia) or multiple basal ganglia and white matter lacunes, or extensive periventricular white-matter lesions.

The distinction between AD+CVD and VaD was also based on the clinical features of patients’ diseases. VaD had a predominant executive dysfunction interfering with daily life activities, and the location and volume of vascular lesions seen on MRI explained the clinical features. AD+CVD, on the other hand, had the clinical features and progression typical of AD, with progressive and prominent episodic memory disorders; executive dysfunction was of secondary importance, and vascular lesions were either an unexpected feature seen on brain imaging or related to vascular risk factors (eg, hypertension or diabetes) or to one or more past strokes and unrelated to the present clinical feature. All diagnoses were revised at each visit. The clinical diagnosis of dementia in our centre was supported by neuropathology in 53 consecutive patients with clinical diagnoses of AD+CVD, VaD and AD who had an autopsy between 1993 and 2007: it was 94.1% for AD+CVD (n = 17), 100.0% for VaD (n = 3) and 97% for AD (n = 33). The autopsy procedures were as follows. The post mortem delay varied from 4 to 30 h. For all patients, the severity of the lesions was estimated by CT-scan and/or MRI and known by the neuropathologist. One hemisphere was frozen for biochemical study of tau proteins, β-amyloid and α-synuclein. The most severely affected hemisphere was exhaustively sampled after formalin fixation for histopathology and immunohistochemistry (τ, β-amyloid, α-synuclein, ubiquitin, prion protein). Whole hemisphere coronal sections kept at the level of the mammilary body were stained with haematoxylin–erythrosin, Bodian silver stain and Kluver–Barrera cyto-architectonic stain. A whole horizontal section of the cerebellum and pons was stained with the same techniques. The elementary lesions indicative of vascular disease were assessed in accordance with recommendations of Jellinger2 at the beginning of the study. All cases were finally reconsidered according to Kalaria et al, as well as the criteria from the International Society of Neuropathology30 31, that is, small-vessel lesions (lacunar infarcts, microlacunae, lipohyalinosis, amyloid angiopathy, microbleeds, myelinic pallor, etc) or large-vessel lesions (territorial infarcts, atheroma of the Willis polygon, etc). Afterwards a neuropathological diagnostic was established matching this evaluation, the τ/amyloid, synuclein burden and the scoring of neurodegenerative lesions. We used the consensus recommendations for post mortem diagnosis of Alzheimer disease (National Institute on Aging and Reagan Institute Working Group on Diagnostic Criteria for the Neuropathological Assessment of Alzheimer’s disease)32 for AD.

Outcome measures

To assess cognitive decline, we compared performances on both the MMSE and the DRS assessed during follow-up. The MMSE was performed at each visit (every 6–12 months). Because the DRS evaluation requires more time to perform, only 482 patients had at least two DRS assessments (and half of them had only two) during the study period. The annual rates of MMSE and DRS decline were calculated as the difference between the first and last MMSE and DRS scores, divided by the time, expressed in years, between the two measurements.

Information on death was routinely collected from relatives or physicians. For patients who could not be contacted or were otherwise lost to follow-up, information was collected through telephone calls to next of kin or physicians.

Other covariates

Age at onset of dementia was determined by informant reporting. Education level was assessed as a qualitative variable in two classes, according to years of schooling: low (L: ⩽8 years) and high (H: >9 years) education levels. Hypertension was defined as a blood pressure value >140/90 mm Hg (or >130/80 mm Hg in patients with diabetes, as recommended by the American Diabetes Association and the Joint National Committee as treatment goals for patients with diabetes), measured at rest on two separate assessments, or current antihypertensive therapy; diabetes mellitus as a fasting blood glucose level >7 mmol/l on two separate occasions, or a current antidiabetic therapy.33 34

Statistical analysis

For simple group comparisons, χ2 tests and ANOVA with post hoc tests to compare multiple groups were used as appropriate. Because we had multiple measurements of MMSE for each patient, MMSE decline was analysed with a multivariate mixed random effects regression model, which takes into account the correlation between repeated MMSE measurements of the same subjects. In addition, this model is robust with missing data and allows analysis of repeated data with variable numbers of measurements per subject, which is the case in our data set. In this model, a significant time effect indicates a change in test scores over time, and a significant interaction between group and time indicates a differential change in the decline of the MMSE score according to type of dementia. To address the non-linearity of cognitive decline, we added an additional adjustment, a time-squared term, in the model and used the Akaike Information Criterion to determine the best-fitting model. In addition, we checked for interactions between time and each of the adjustment variables. MMSE decline was studied from baseline evaluation to a maximum of 4 years, to ensure sufficient MMSE evaluations in each dementia group. In the model, the variable “time” was studied as number of months, but fig 1 presents the means of predicted adjusted values for 6-month periods, to maximise its clarity. Because we did not have enough DRS measurements to conduct the same analysis for the DRS decline, we used a multivariate linear regression model to test the association between the mean annual DRS decline and dementia type. Multivariate analyses of cognitive decline were adjusted for age, sex, education, hypertension, diabetes, presence of an informant, and baseline MMSE and DRS when appropriate.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Adjusted predicted Mini Mental State Examination (MMSE) decline according to type of dementia (N = 970; mixed random effect regression model). Results adjusted for age, sex, education, hypertension, diabetes, presence of an informant, and baseline MMSE and Dementia Rating Scale. AD, Alzheimer disease; CVD; cerebrovascular disease; VaD, vascular dementia.

As we were interested in the progression of the disease, the time elapsed on the abscissa for the survival analysis is the time since the onset of dementia (ie, duration of disease) rather than since study entry. To take into account the fact that patients with known disease onset may die before entering the study, an adjustment known as “delayed entry” or “left truncation” was used.35 The Cox proportional hazard model, which allows this delayed entry, was used to assess the relative mortality risk (RR) by type of dementia, and actuarial analysis was used to predict survival curves for them. Data were right-censored 10 years after disease onset. Survival models were adjusted for age, sex, educational levels, diabetes, hypertension, presence of an informant, and baseline MMSE and DRS. The proportionality assumption was tested and fulfilled for all of the variables studied by introducing a time-dependent coefficient in the Cox proportional hazards models. SAS 8.02 (SAS Institute, Cary, North Carolina) software was used for the statistical analyses.

RESULTS

Patients’ characteristics at baseline and during follow-up are reported in the table by type of dementia. Men had VaD more often than AD or AD+CVD (p<0.0001). AD+CVD patients were older than AD or VaD patients at onset and at first visit (post hoc test: AD+CVD vs AD: p<0.01, AD+CVD vs VaD, p<0.0003, AD vs VaD: p = NS). Baseline MMSE (post hoc test: AD+CVD vs AD: p = NS, AD+CVD vs VaD and AD vs VaD: p<0.0003) and DRS (post hoc test: AD+CVD vs AD: p = NS, AD+CVD vs VaD and AD vs VaD, p<0.004) were highest for VaD compared with AD+CVD or AD patients. As expected, most patients with AD and AD+CVD received AChEI treatment (95%) and patients with VaD and AD+CVD had hypertension, a stroke history and diabetes more often than AD patients (p<0.0001).

Cognitive decline

The mean annual MMSE decline differed significantly according to dementia type: 2.0 (2.5) points for patients with AD alone, 1.5 (2.3) points for AD+CVD and 0.6 (2.7) points for VaD (table 1, post hoc tests: AD+CVD vs AD, AD+CVD vs VaD and AD vs VaD: p<0.05). As expected, MMSE scores declined significantly over time (p<0.0001) in the adjusted mixed model and, as in the univariate analysis, the decline rate was statistically different between dementia types (test for interaction time and dementia, p = 0.03). Decline was fastest for AD compared with VaD patients (β estimate: −0.08 (0.03), p = 0.01) and did not differ significantly between patients who had AD+CVD and those with either VaD (β estimate: −0.06 (0.04), p = 0.16) or AD (β estimate: 0.02 (0.03), p = 0.65) (fig 1). Introduction of a time-squared term did not improve the goodness of fit of the model. The mean annual DRS decline differed significantly according to dementia type: 8 (12) points for AD patients, 5 (11) points for those who had AD+CVD and 4 (6) points for VaD patients (post hoc tests: AD vs VaD and AD+CVD vs AD: p<0.05, AD+CVD vs VaD: p = NS, table). This result did not change after adjustment for age, sex, education, diabetes, hypertension, presence of an informant and baseline MMSE.

Survival

AD patients had a longer follow-up than the other patients (table 1, post hoc test: AD+CVD vs AD: p = 0.07, AD+CVD vs VaD, p = NS and AD vs VaD: p = 0.04). Of patients who died within 10 years of disease onset, the duration of disease was similar for all three dementia groups (VaD: 6.1 (2.1), AD+CVD: 6.8 (1.8), and AD: 6.8 (1.8) years, p = NS). Figure 2 shows adjusted actuarial survival curves for the three dementia groups. After adjustment, risk of mortality was similar for VaD (RR 0.7 95% CI (0.5 to 1.1), p = 0.2) compared with AD patients. Patients who had AD+CVD tended to have a lower risk of mortality; although the difference was not significant (RR 0.7 95% CI (0.5 to 1.0), p = 0.06). Risk of mortality according to dementia remained unchanged without adjustment for age. The older the patients at clinical assessment, the sooner they died, independently of diagnosis and baseline cognitive performances (RR = 1.1 95% CI (1.0 to 1.1) for each year of age, p<0.0001).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Actuarial survival curves from diagnosis according to type of dementia (N = 970). Results adjusted for age, sex, education, hypertension, diabetes, presence of an informant, and baseline Mini Mental State Examination and Dementia Rating Scale. AD, Alzheimer disease; CVD; cerebrovascular disease; VaD, vascular dementia.

Regardless of diagnosis, the shorter the time between first symptoms and first visit, the longer the patients survived (RR = 0.7 (0.6 to 0.7), p<0.0001 for each year earlier the first visit occurred). Risk of death was lower for women (RR = 0.4; 95% CI (0.3 to 0.6), p<0.0001). When we estimated survival from disease diagnosis instead of disease onset, results were unchanged.

DISCUSSION

In this clinic-based population study comparing patients with AD (n = 663), VaD (n = 141) and AD+CVD (n = 166) for global cognitive decline and survival rate (mean follow-up: 4.7 (2.7) years), we found significant differences between patients with different types of dementia. Patients who had AD+CVD were older than AD or VaD patients at both onset and first visit, and the latter two were similar for both. Cognitive decline, assessed with both MMSE and DRS scores, was significantly faster for AD than for VaD patients. The rate of decline was essentially intermediate for those who had AD+CVD, although this trend was not significant. Mortality risk did not differ between VaD and AD patients, and tended to be lower for those who had AD+CVD compared with AD patients.

The characteristics of our population differ from those of most other studies: they were somewhat younger (65–85 years old) than in most population-based studies,5–7 36–46 but within the range reported by other clinic-based studies (60–78 years old).4 7 8 10 22 25 47^–49 The higher age for patients who have AD+CVD compared with AD can be explained by the increase in cerebrovascular lesions visible in MRI, and it is these lesions that differentiate pure AD from AD+CVD.15 This probably explains the difference in the prevalence of AD+CVD that we found compared with population-based studies. Because population-based studies do not systematically perform MRI, their proportion of so-called AD+CVD dementia is lower (3–10%) than in clinic-based studies.45 The higher rate of men among VaD patients compared with AD+CVD and AD patients is consistent with other studies.4 7 8 This result may be due to differences in biological characteristics, or in survival rates or in exposure to protective or risk factors among dementia types.

Our results about survival according to dementia type differ from those of other studies: 11 studies report better survival rates for AD than VaD patients6 7 38–40 42–44 46 49 and three no difference in survival according to dementia type.4 36 45 Our results about better survival for AD+CVD are surprising. Because these patients were older than the others, we would have expected a lower survival for them. To attempt to understand this result better, we estimated the RR of death without adjusting for age. We expected the RR would be higher for these patients than for those of AD. Results were unchanged, however.

Survival time for our patients was longer than in recent population-based studies.5 7 Our patients were followed and treated in a memory clinic (95% of our AD+CVD and AD patients took AChEI), which may explain in part their longer survival, regardless of dementia type, compared with population-based patients. In addition, some patients with rapidly progressive dementia may die before consulting in a memory clinic and are thus excluded from our sample. Our delayed entry survival model corrects for this issue, but probably not entirely.5 Surprisingly, our survival time was also longer than in some recent clinic-based studies.4 6 8 Selection of our population may explain this observation: our patients were younger and had hypertension less often than in those studies, and risk of death increases with age for demented patients.4 6^–8

Nevertheless, we also found many standard prognostic factors in our sample, such as the higher risk of death among men.6 Our cohort was also newer than most of the clinic-based studies. Thus, changes in practices may also explain the longer survival we found compared with the literature. Nowadays, patients consult earlier after their first symptoms, and we found that the shorter the delay between first symptoms and first visit, the longer the survival. Moreover, better control of vascular risk factors for VaD patients may explain this change, at least in part. Thus, VaD patients who take aspirin or antiplatelet treatment have a lower risk of death than others.4 50 These patients may also have used other more recent treatments, such as statins, which have proven to protect against cerebrovascular and coronary events in patients at high risk of vascular disease.51 52 Furthermore, as our team has previously reported, lipid-lowering agents are associated with slower cognitive decline in degenerative dementias such as AD.53 Nevertheless, these survival results cannot be generalised to the general population; instead they reflect recent clinical practices for patients who seek care.

The annual MMSE decline of 2.0 (2.5) points for our AD patients appears to fall within the range of recent studies.10 11 37 It is nevertheless important to note that this annual decline has been decreasing since the 1990s and the introduction of AChEI treatments (the annual MMSE decline at that time ranged from 3.0 to 4.5 points per year).36 The earlier recognition of dementia may account for some of the decreased MMSE decline. The annual MMSE decline of 0.6 (2.7) points for VaD patients is lower than that reported in other studies.10 11 This lower cognitive decline is however not surprising and has been described in some studies, because the decline in VaD patients is more likely to affect executive functions, which are estimated instead with an activity of daily living scale.54 55 Moreover, executive dysfunctions are the first signs leading to a diagnosis of dementia in VaD, and this may explain why our VaD patients had a higher baseline MMSE than AD or AD+CVD patients. Only one study with a sufficient number of cases found no difference in the rate of cognitive decline for patients with VaD and AD.36 A recent publication reports results similar to ours, with a lower decline for VaD patients during follow-up.11 Decline in patients who had AD+CVD was intermediate between those with AD and with VaD, probably because of the coexistence of both AD and cerebrovascular disease.

As pointed out earlier, the different criteria used to define VaD and AD+CVD are likely to account for a large part of these discrepancies. Thus, misclassification between AD+CVD, AD and VaD is common, and as Knopman showed, different criteria for VaD influence the survival results greatly.6 Neuroimaging, required for the NINDS-AIREN criteria, is essential to confirm the presence of “probable” VaD. Moreover, neuropathological evaluation indicates that the NINDS-AIREN criteria appear to be the most specific of the frequently used VaD criteria.56 AD+CVD is diagnosed by the presence of pathology of both AD and VaD, but its specific definition remains controversial, and some authors have redefined its criteria.2 It is possible today, however, to distinguish accurately between typical AD and typical VaD, and mixed or unclear cases should be studied separately.12 Nevertheless, much of the literature does not distinguish between AD+CVD and VaD or AD.45

The diagnostic procedures used in our memory clinic should have minimised diagnostic misclassification. First, all patients had a neuroimaging examination so that we were able to restrict our analysis to “probable” cases of VaD and therefore to exclude “possible” cases of VaD, the diagnosis of which is more likely to change as it progresses. Second, we separated AD+CVD from typical AD and from VaD, as recently recommended.12 Although we have no pathology data for the dementia type of most patients in this series, the accuracy of the clinical diagnosis of dementia in our centre is good. Finally, there is no consensus about the method for assessing onset of dementia. The use of informant reports, as in our study, has been compared with reports from medical records.57 Conclusions of this study show that informant reports of age of dementia onset are reliable and reasonably valid, even when only a single question is asked as part of a larger interview.

Higher age at diagnosis is associated with shorter survival, and our patients were younger than those in most studies.5–7 36–46 Thus, the differences in age at diagnosis may also account for a part of the discordance between our findings and those of other studies of survival and cognitive decline. Furthermore, our patients were recruited more recently than in most previous publications, and more recent publications have tended to find results similar to ours.4 45

Our population of demented patients seen in a tertiary care centre does not represent population-based demented patients, but their study allows us to describe the characteristics and course of demented patients seen in a memory clinic. The three dementia groups came from the same memory centre and underwent the same selection procedure and medical care. On these points, our groups are probably comparable. Moreover, clinicians need to have data representative of their patient populations. These data can help them to adjust care more appropriately and to inform families about the progression of the dementia. Some results are interesting from this point of view, in particular, the inverse correlation between time between first symptoms and first visit and survival after first symptoms. This suggests the beneficial role of early healthcare. Regularly updated data about the age at first consultation or the percentage of treated patients is also useful for clinicians.

Because dementia is today a major public health problem for which neither preventive nor curative treatments have yet been discovered, regularly updated data describing the characteristics and progression of different types of dementia treated in a memory clinic are needed. This clinical cohort study shows that patients with VaD, AD and AD+CVD present different baseline and follow-up characteristics, and underlines the necessity of distinguishing VaD from both AD and AD+CVD.