Introduction

The last few years have seen considerable progress in revisions of clinical diagnostic criteria for dementia. In relation to Alzheimer's disease (AD), the focus of this particular debate, the original McKhann criteria were described in 1984 (McKhann et al., 1984) and represented a landmark in their field, allowing accurate categorisation of a type of dementia that was indeed shown subsequently by validation studies to largely represent AD. The criteria distinguished between “probable” AD, which despite its name was the highest level of clinical certainty available, “possible” AD, when there were variations in the course and presentation or the presence of another comorbidity that could be the cause of dementia, and “definite” AD for which autopsy validation was necessary. The impact of these criteria cannot be overestimated. Most of what we know regarding the natural history, and certainly the therapeutic response (and sadly non-response) of AD, is based upon cohorts diagnosed with these criteria, and most usually the probable Alzheimer type. Validation studies have shown reasonably good, and clinically applicable, sensitivity for the criteria of around 80%, but specificity has been lower, at around 70% (Knopman et al., 2001). This low specificity is an issue, largely because of the lack of any positive features for ruling AD in (they have largely, though somewhat unfairly, been labelled criteria of exclusion). In addition, the low specificity is in part because the criteria often also include conditions that were not well defined or even recognised in the 1980s as separate disorders. For example, dementia with Lewy bodies, subcortical ischaemic vascular dementia, and various subtypes of frontotemporal dementia, which have now been shown to have distinct clinical, pathological and, in some cases, genetic components.