Diagnostic accuracy of a two-cut-off approach using the FAQ/MMSE ratio and FAQ for clinical preselection of patients for anti-amyloid therapy.

Abstract

Anti-amyloid therapies for Alzheimer's disease (AD) require efficient patient selection. The Clinical Dementia Rating (CDR) scale is the reference standard for staging, but it is time-consuming to administer. Simple tools to distinguish early-stage cognitive impairment (CDR 0.5-1) from more advanced stages (CDR 2-3) would therefore be of substantial clinical value. Participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohorts 1-3 with CDR ≥0.5 were analysed. Three logistic regression models using Mini-Mental State Examination (MMSE), Functional Assessment Questionnaire (FAQ) or the FAQ/MMSE ratio as predictors were developed. Two cut-offs per model were selected to ensure minimum sensitivity and specificity of 0.99, defining rule-out, rule-in and intermediate (uncertain) zones. Performance was assessed using discrimination, calibration and decision curve analysis. Among 1533 ADNI participants with CDR ≥0.5, two-thirds (n=1022) were assigned to the training set and one-third (n=511) to the test set. FAQ/MMSE and FAQ showed excellent discrimination (area under the curve, AUC 0.97-0.98), outperforming MMSE (AUC 0.94-0.95). FAQ/MMSE demonstrated the best overall performance, although differences compared with FAQ were small and not statistically significant. Dual cut-offs for FAQ/MMSE (0.67 and 1.44) and FAQ (12 and 27) enabled clinically meaningful stratification, with 80% and 70% of participants classified into high-confidence zones, respectively. Results were consistent across the training and test sets. The FAQ/MMSE ratio and FAQ score show high accuracy in distinguishing early-stage (CDR 0.5-1) from more advanced cognitive impairment (CDR 2-3). These simple measures may support the clinical preselection of patients for further evaluation in the context of anti-amyloid therapy.