Development of a Screening Algorithm for Alzheimer’s Disease Using Categorical Fluency and Confrontational Naming Abilities

Development of a Screening Algorithm for Alzheimer’s Disease Using Categorical Fluency and Confrontational Naming Abilities

Objective：Declines in naming ability and semantic memory are well-known features of early Alzheimer’s disease (AD). We developed a new screening algorithm for AD using two brief language tests : the Categorical Fluency Test (CFT) and 15-item Boston Naming Test (BNT15). Methods：We administered the CFT, BNT15, and Mini-Mental State Examination (MMSE) to 150 AD patients with a Clinical Dementia Rating of 0.5 or 1 and to their age- and gender-matched cognitively normal controls. We developed a composite score for screening AD (LANGuage Composite score, LANG-C) that comprised demographic characteristics, BNT15 subindices, and CFT subindices. We compared the diagnostic accuracies of the LANG-C and MMSE using receiver operating curve analysis. Results：The LANG-C was calculated using the logit of test scores weighted by their coefficients from forward stepwise logistic regression models : logit (case)=12.608-0.107×age+1.111×gender+0.089×education-0.314×HS 1st -0.362×HS 2nd +0.455×per- severation+1.329×HFCR 2nd -0.489×MFCR 1st -0.565×LFCR 3rd . The area under the curve of the LANG-C for diagnosing AD was good (0.894, 95% confidence interval=0.853-0.926 ; sensitivity=0.787, specificity=0.840), although it was smaller than that of the MMSE. Conclusion：The LANG-C, which is easy to automate using PC or smart devices and to deliver widely via internet, can be a good alternative for screening AD to MMSE.