Clinical diagnosis of presenile Alzheimer's disease: A novel approach

Abstract

A national retrospective survey of hospital records was used to select those patients with presenile dementia who had undergone neuropathological examination. The National Institute of Neurological and Communicative Disorders and the Alzheimer's Disease and Related Disorders Association (NINCDS–ADRDA) clinical criteria for probable Alzheimer's disease and a Hachinski score were applied to each record before the neuropathological diagnosis was known. A discriminant analysis, which entered the criteria as variables, was performed. The diagnostic accuracy of the clinical criteria was compared before and after discriminant analysis. One thousand six hundred and seventy-one records were scrutinised of which 61 patients had undergone neuropathological examination. NINCDS–ADRDA criteria had a diagnostic accuracy of 72% (specificity 88%, sensitivity 61%) compared to 77% (specificity 80%, sensitivity 75%) after analysis. NINCDS–ADRDA criteria together with the Hachinski score had an accuracy of 72% (specificity 61%, sensitivity 88%) compared to 83.6% (specificity 76%, sensitivity 89%) after analysis. Variables of highest discriminating value were the Hachinski score, presence of coexistent neurological disease and presence of coexistent systemic disease. The results highlight limitations of current clinical criteria used to diagnose Alzheimer's disease and suggest that substantial improvements are possible.