eventually leading to the clinical syndrome of dementia, may start as early as decades before the disease becomes clinically overt. In analogy with the gradually accumulating neuropathology, the transition from healthy to demented is also gradual, rather than abrupt. The moment when dementia is diagnosed is in fact arbitrary. Moreover, the artificial dichotomisation between healthy and demented does not do justice to the continuum of cognitive (dys)function. The concept of mild cognitive impairment (MCI) has been developed to account for the transitional phase between healthy and demented.15 However, introducing concepts like MCI only shifts the problem, as the borders between healthy and MCI and between MCI and demented remain just as arbitrary and unclear. A possible solution would be to discard the arbitrary distinction between normal and demented, and instead use a continuous outcome, such as a test of cognitive function. This would have several advantages. First, costs and time can be saved as the extensive diagnostic work-up is not necessary anymore. Second, by abolishing the artificial dichotomisation into normal and demented, the continuum of cognitive decline is done more justice. This approach also provides the opportunity to study progression of decline within demented individuals. Biomarkers A third issue reflects the complex relationship between the syndrome of dementia and the underlying diseases. When we talk of AD, we refer to the syndrome that is characterised by progressive memory problems, which usually has an insidious onset, etc. However, at the moment the diagnosis of AD is made, we assume to know the underlying neuropathological substrate—that is, neuritic plaques and neurofibrillary tangles. We assume to know this, because during lifetime, it is impossible to directly measure neuropathology. In fact, postmortem studies have shown that this assumption in many cases is wrong.16 In a report of the MRC CFAS of the first 209 subjects (48% demented) who came to necropsy, Alzheimertype pathology and vascular pathology were equally common, and both correlated with cognitive decline. Most subjects had mixed pathology. Approximately one third of clinically demented patients did not fulfil neuropathological criteria for definite AD, whereas an equally large proportion of nondemented elderly subjects did fulfil these criteria.16 Neuropathologically, the distinction between different types of dementia, and even between demented and non-demented, seems to be very difficult. The question arises that if it is useful to make clinical distinctions between subtypes of dementia, neuropathology may not even exist. A step towards directly measuring disease, rather than clinical phenotype, would be to take biomarkers as outcome of studies. Both neuroimaging and cerebrospinal fluid can provide useful surrogate markers that give a more direct impression of the pathology. In this way, the possibility of different types of pathology coexisting within one subject is appreciated. For example, magnetic resonance imaging (MRI) measures suggestive of Alzheimer-type pathology and vascular pathology can be evaluated simultaneously. Cross sectional versus longitudinal studies Studies with a longitudinal design are preferred over studies with a cross sectional design for several reasons. It is conceivable that information about risk factors may be systematically different between patients and controls. Patient data must come from a proxy, who might recall the medical history differently than a proxy of a control or the control himself. In addition, prevalence is determined by both the number of new cases over a given period of time, and by the duration of survival once patients have the disease. In analogy, findings of cross sectional studies can reflect the contribution a risk factor makes to developing dementia as well as to surviving after the dementia starts. v6 NEUROLOGY IN PRACTICE www.jnnp.com copyright. on December 4, 2021 by guest. Protected by http://jnnp.bmj.com/ J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.2005.082867 on 16 November 2005. Downloaded from Another important issue in this