Compare the performance of category fluency to that of letter fluency, MoCA (Montreal Cognitive Assessment), and MMSE (Mini-Mental Status Exam) in detecting mild cognitive impairment (MCI).  

The MMSE and MoCA are brief cognitive screens used to detect MCI. Letter fluency is included in the English MoCA, but not category fluency, despite a tendency for the latter to decline early in Alzheimer’s disease. Limited sample sizes in the existing literature contribute to uncertainty regarding use of category fluency as a comparable screen.  

We extracted initial study-visit data from the National Alzheimer’s Coordinating Center (NACC) database for two measures of category fluency  (n=32762), two measures of phonemic fluency  (n=13731), the MMSE (n=18929), and the MoCA (n=13833) for participants clinically diagnosed as either cognitively unimpaired (CU) or MCI.  We performed ROC (Receiver Operating Curve) analyses to detect MCI from non-MCI (normal and insufficient impairment for MCI) per clinician diagnosis at the initial visit, based on these measures. Statistical comparisons were based on bootstrapping ROC Area Under the Curve (AUC) values, with n = 20,000 iterations. Claude Sonnet 4.0 was used for assistance with MATLAB code generation for data analysis. 

Both category fluency measures, Animal fluency (AUC=0.72, Accuracy=0.703) and Vegetable fluency (AUC=0.73, Accuracy=0.708), displayed significantly higher AUC and accuracy levels than both phonemic fluency measures, fluency for the letters F (AUC=0.64, Accuracy=0.681) and L (AUC=0.63, Accuracy=0.682), for detecting MCI vs CU individuals (p<0.0001 in all cases). Combining animal and vegetable fluency further increased diagnostic accuracy (AUC=0.75, Accuracy=0.722, p<0.0001), reaching levels at least approximate to the MMSE (AUC=0.73, Accuracy=0.723), while the MoCA was superior (AUC=0.79, Accuracy=0.752).