Diagnostic criteria for vascular dementia: a step towards new ones?

Vascular dementia (VaD) is a complex concept of particular interest; first, because it is a frequent and potentially preventable cause of dementia and second, because of its link to Alzheimer's disease (AD). In the early twentieth century, atherosclerosis was considered as the main cause of dementia in the elderly and Alzheimer's disease was considered as a rare form of presenile dementia. The first description of VaD was made by Mayer-Gross in 1969.1 He first pointed out the importance of arterial hypertension, which he reported to be the main cause of dementia in approximately 50% of all VaD cases. There are many variations on the VaD concept. The concept of ‘multi-infarct dementia’ (MID) was introduced by Hachinski et al. in 1974.2 In 1983, Loeb and Gandolfo proposed the broader ‘vascular dementia’ concept to take into account the multiplicity of the etiologies.3 In 1995, Bowler and Hachinski pointed out the fact that many patients with cerebrovascular disease do not have dementia, but are only ’cognitively impaired’, thus defining a new concept: ‘vascular cognitive impairment’.4 Erkinjuntti, in 2002, proposed that ‘subcortical vascular dementia’ should be considered as the archetype of VaD.5 The diagnosis of VaD involves two steps: Diagnosis of dementia per se and establishment of its vascular etiology. While the first step is quite well defined, the second step is quite tricky as neither significant vascular etiology nor a strict ‘gold’ standard for diagnosis has been defined. Correlations between clinical diagnosis and autopsy remain weak. In 1997, Snowdon et al. showed, in an autopsy study, that the frequency of minor cerebrovascular pathology is rather important in healthy elderly people and in association with other forms of dementia, mainly AD.6 In published dementia autopsy reports, the probability of occurrence of VaD may vary from 9%7 to 70%.8 The frequency of VaD (mixed and pure) is 7% for women and 10% for men before the age of 70 years, and reaches 23%4, 5 for women and 20% for men after the age of 70 years.9 The accuracy of clinical diagnosis as compared with autopsy results is also quite poor. The sensitivity varies from 45%10 to 83%.11 Given the absence of a pathological ‘gold standard’, it is no longer possible to rely on magnetic resonance imaging (MRI) studies. Recently, Erkinjuntti et al. has stressed the need for precise guidelines for determining the significance of cerebrovascular lesions observed using MRI.12 At the present time, it is impossible to evaluate the impact of cortical or subcortical lesions. White matter lesions are supposed to be a very important load on cognitive deficit, but there is no precise guideline as to which type of lesion should be considered as a contributing lesion. There is only a ‘soft consensus’ on the recommendation that such a lesion should be diffuse, more than 25% extensive, and of irregular shape. It is very strange that quantitative assessment techniques, widely employed in multiple sclerosis studies, have not been systematically utilized in VaD studies yet. Another source of difficulty comes from the multiplicity of VaD etiologies: single or multiple infarct; non-infarction ischemia, mainly subcortical leukoencephalopathy; or hemorrhages. Therefore, the signs and symptoms are mostly variable, depending on lesion type, side and site. For instance, dementia resulting from large-vessel multiple infarctions presents variable associated cortical signs, depending on the site of the lesion. In lacunar disease, which develops because of the damage to small subcortical perforating vessels, frontocortical signs are often prominent, as a result of the interruption of frontal subcortical loops. Brain imaging is also quite heterogeneous and the relationship between imaging and clinical signs is controversial. Damage to white matter could remain undetected until it reaches more than 6% of the volume. By then, the symptoms may be progressive and rather similar to those seen in AD. Tomlinson has tried to define VaD using a damage volume threshold of more than 100 mL,13 but Loeb et al. has stressed that it could be more a matter of the lesion location: Damage to some crucial strategic sites within the cortex, basal ganglia and white matter could be of greater importance than the volume of damage.14 Another very important issue is the cognitive profile. Based on the variability of lesion sites, memory can be preserved in VaD, while it is always impaired in AD.15 Therefore, the presence of a memory impairment should not be a mandatory criterion for the diagnosis of VaD (contrary to the diagnosis of AD). Unfortunately, it is easy to verify that almost all the criteria for VaD diagnosis specify the presence of a memory impairment. A short critical review of existing diagnostic criteria for VaD will show that most criteria rely on the concept of multiple infarct dementia. The criteria specified in the following are based on expert consensus, but not on data: Hachinski ischemia score (HIS), Diagnostic and Statistical Manual of Mental Disorders, revised third edition (DSM-III-R), Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV), International Statistical Classification of Diseases and Health-related Problems, 10th revision (ICD 10), State of California Alzheimer's Disease Diagnostic and Treatment Centers (ADDTC) and National Institute of Neurological Disorders and Stroke–Association Internationale pour la Recherche et l’Enseignement en Neurosciences (NINDS–AIREN). Wetterling et al. have shown that only three items in the HIS differentiate between VaD and the other types of dementia: history of stroke, focal neurological symptoms and focal neurological signs.16 HIS is particularly poor at differentiating between VaD and mixed dementia (sensitivity = 93%; specificity = 17%) and between mixed dementia and AD (sensitivity = 84%; specificity = 29%).17 The definition of dementia in the DSM-III-R is based on the concept of AD, which requires the impairment of short-term and long-term memory. As explained previously, this impairment is not constant in VaD. In a study of 167 patients, only 10 patients fulfilled the criteria (10.2%).16 Thus, by using the DSM-III-R criteria, a physician would introduce a bias that favors the representation of VaD patients with memory impairment, thus making it very difficult to distinguish between such patients and AD patients on clinical grounds. At first glance, DSM-IV seems to constitute a better set of criteria. However, it has some shortcomings. The DSM-IV definition is rather broad, based mainly on the MID concept and is also derived from the AD profile. There are no guidelines for clinical signs and for brain-imaging requirements. The ICD 10 definition is very close to that in the DSM-IV, although it does state to its merit that the pattern of cognitive impairment may be quite variable. In the study of 167 patients by Wetterling et al., only 31 (31.6%) patients fulfilled all the criteria, showing that the ICD 10 definition is not very efficient.16 The ICD 10 definition is rather selective, but it is mainly based on the MID concept and lacks detailed guidelines for clinical signs, brain-imaging requirements and etiological analysis. The ADDTC,18 which is exclusively devoted to the diagnosis of ischemic VaD, and NINDS–AIREN19 provide more recent and improved criteria that are better focused on the connection between dementia and vascular damage. Both the ADDTC and NINDS–AIREN definitions require the presence of dementia, cerebrovascular disease and a time relationship between the two. They express different levels of certainty (probable, possible, definite) and recognize the heterogeneity of VaD. However, their inter-rater reliability is moderate to fair, with a kappa value between 0.46 and 0.72.20 Both definitions also have some weaknesses. The ADDTC definition requires one computed tomography (CT) or T1-MRI infarct outside the cerebellum, but white matter lesions do not qualify for the definition. The NINDS–AIREN definition requires either multiple infarcts, extensive white matter lesions or a strategic single infarct. When comparing the effectiveness of these various VaD criteria, they appear to identify different clusters of patients. Wetterling et al. have shown that within a single sample, the estimated prevalence of VaD may vary by a factor ranging from one to five: 52% with DSM-IV; 33% with ICD 10; 27% with ADDTC; and 14% with NINDS–AIREN.16 Similar results have been obtained in other studies.21-24 Comparison between the NINDS–AIREN and ADDTC results22 in a neuropathological study of 103 patients with dementia showed that the sensitivity of ADDTC criteria is higher (63% versus 58%), but its specificity is lower (64% versus 80%). About 29% of patients with mixed dementia are misclassified using NINDS–AIREN criteria and 54% using ADDTC criteria. Thus, the present criteria are inappropriate for identification of mixed dementia. Therefore, it is necessary to define new VaD criteria, based on data and ‘not on dogma’.25 Studies should be conducted on more homogeneous subtypes, such as subcortical VaD, as suggested by Erkinjuntti et al.12 The concept of mixed dementia has to be refined in terms of cognitive profile and brain-imaging requirements. The main challenge now is to find the best study setting to develop those criteria. Bowler25 wrote that ‘a vascular clinic would otherwise be a more appropriate forum than a memory clinic’. It may be that the VaD criteria are inappropriate because they have been developed through memory clinics, more for the exclusion of VaD patients from AD studies rather than specifically for the analysis of vascular dementia.