Comparative diagnostic accuracy and clinical utility of the Mini-Addenbrooke’s Cognitive Examination, ACE-III, and MMSE for functional staging across the Alzheimer’s disease continuum

Introduction. The advent of disease-modifying therapies has heightened the demand for cognitive screening instruments capable of detecting cognitive impairment and supporting both functional staging and longitudinal monitoring of Alzheimer’s disease (AD). Although the Addenbrooke’s Cognitive Examination III (ACE-III) offers a comprehensive multidomain assessment, the Mini-Addenbrooke’s Cognitive Examination (M-ACE) may be more practical for routine clinical use. However, the extent to which M-ACE supports disease staging remains insufficiently established. Objective. To compare the diagnostic accuracy and functional staging capabilities of M-ACE, ACE-III, and the Mini-Mental State Examination (MMSE) across the AD continuum. Material and methods. This retrospective study included 1,834 consecutive individuals seeking clinical evaluation, classified as having subjective cognitive decline (Functional Assessment Staging Tool [FAST] stage 2), mild cognitive impairment (FAST stage 3), or AD dementia (FAST stages 4–6). Analyses included assessments of internal consistency, convergent validity, receiver operating characteristic (ROC) analyses, determination of optimal cutoff values, and multivariable logistic regression. Diagnostic accuracy was compared using DeLong’s test. Results. M-ACE exhibited good internal consistency (McDonald’s ω = 0.835; Cronbach’s α = 0.731) and strong convergent validity with ACE-III (ρ = 0.822). ACE-III demonstrated the most consistently high diagnostic accuracy across FAST transitions. M-ACE performed nearly as well as ACE-III at earlier stages and significantly outperformed MMSE at the FAST 2–3 transition, whereas its discriminative performance declined at later stages. MMSE yielded the highest area under the receiver operating characteristic curve (AUC) at FAST 4–5, although its performance did not differ significantly from ACE-III. Optimal M-ACE cutoff values were ≤ 25 (FAST 2–3), ≤ 19 (FAST 3–4), ≤ 13 (FAST 4–5), and ≤ 7 (FAST 5–6). M-ACE remained independently associated with disease stage after adjusting for age, sex, and education. Exploratory analyses revealed that Memory Recall exhibited a floor effect from FAST stage 4, accounting for the increasing advantage of ACE-III in later disease stages. Conclusions. M-ACE offers an optimal balance between diagnostic accuracy and administration time, positioning it as a practical alternative to MMSE for routine cognitive screening and longitudinal monitoring. The findings support ACE-III as the preferred instrument for comprehensive baseline characterization, while M-ACE is particularly suitable for rapid follow-up assessments.