Retrospective evaluation of revised criteria for the diagnosis of Alzheimer's disease using a cohort with post-mortem diagnosis

Objective: The criteria currently used to diagnose Alzheimer's disease (AD) require the presence of dementia, i.e. cognitive impairment sufficient to affect normal social and/or occupational function. Dubois et al. (Dubois et al., 2007) have recently proposed a set of revised criteria that may aid the diagnosis of the earlier stages of AD, and do not require the presence of dementia. We aimed to evaluate the new predementia-AD criteria through their retrospective application to the OPTIMA cohort with post-mortem (PM) confirmed diagnoses. Methods: The criteria were evaluated for sensitivity and specificity using cognitive, neuroimaging and cerebrospinal fluid data and clinical information for exclusion criteria. Limitations in choice of cognitive test, use of CT scans rather than M RI and missing CSFs affected the outcomes. Analyses were carried out for the whole cohort (n = 243) and on a mild-stage subgroup (n = 99) defined by MMSE >= 21. Results: Of the four options for fulfilling the revised-criteria, the best results for the whole cohort were achieved using memory and CSF data with exclusion criteria applied (.68 sensitivity and .93 specificity). The pattern was similar for the mild cohort, but with lower sensitivity. Specificities of 1.0 were reached with supportive criteria, CSF and CST plus MTL. Conclusions: The revised-criteria, when applied to our cohort, offer good specificity and reasonable sensitivity when compared with the gold standard of PM diagnosis. The criteria were not more effective for early stage dementia. Copyright 2010 John Wiley & Sons, Ltd.