Objective: Recent advances in treating Alzheimer's disease (AD) have shifted clinical focus to identification of patients in a mild disease stage. Clinical tests occasionally used for detecting early AD, such as the Mini-Mental State Exam (MMSE), are too long for routine screening. Experimental tests such as recall measurement and category fluency have been shown to be the most effective instruments for distinguishing normal individuals from early AD patients. Now, a very brief and highly efficient screening test is needed for clinical recognition of mildly impaired patients likely to have early AD. This study presents an approach to the development of a short test to screen for AD.
Design: To develop a Brief Alzheimer Screen (BAS), characteristics of MMSE items and category fluency were analyzed using a derivation and validation sequence. Data from CERAD (Consortium to Establish a Registry for Alzheimer's Disease), consisting of 406 normal individuals and 342 patients with mild AD (MMSE>19), were used as a representative population. Both normal and mild AD patient samples were randomly divided into derivation and validation subgroups. Logistic regression was performed on the derivation subgroups to determine a predictive equation that would best separate the normal and mild AD groups.
Results: The resulting model for discriminating between these groups included items in the following order: recall (R, 3 points), animals (# in 30 secs, A), date (D, 1 pt.), spell (WORLD backwards, S, 5 pts), (all p<0.0001). All other MMSE items contributed a negligible portion of the variance. Logistic regression analysis for identifying AD or control group membership using these most significant variables produced the following equation (r2=0.77):
BAS = 3 x R + 2/3 x A + 5 x D + 2 x S
This equation was applied to the validation subgroup to estimate sensitivity and specificity. From the results, dual cut-points are recommended for screening. For scores of 22 and below, sensitivity was over 98% with specificity of 87% or less. For a score of 27 or greater, the sensitivity was 90% or less and specificity was over 99%.
Conclusion: These data support the BAS as adequately powerful for use in screening for patients over 60 years of age for cognitive impairment that could represent early AD.
Back to PB Tuesday Poster Sessions
Back to The Eleventh International Congress