Abstract
Introduction: To identify the optimal A PET threshold for distinguishing fast progressors to Alzheimer's disease (AD) in individuals with mild cognitive impairment (MCI) and to assess whether a cutoff based on Mini-Mental State Examination (MMSE) performance can improve or substitute for the A PET cutoff. This research will help refine risk stratification in MCI and guide clinical decision-making regarding early therapeutic interventions.
Methods: We included 686 MCI participants with Clinical Dementia Rating (CDR) score of 0.5 from two cohorts, followed for up to 7 years. Harmonized data from AIBL (N=166) and ADNI (N=520) were analysed using Cox proportional-hazards models, adjusted for age, sex, and APOE4 status, with the event of interest being progression to mild dementia due to AD (detected by CDR = 1 and -amyloid positivity (A >20 CL)). Optimal thresholds for MMSE (27) and A (44 CL) were selected to maximize hazard ratios (HR) at 3 years, categorizing participants into low-risk and high-risk groups based on cognitive performance (187 participants, 27%, MMSE<27 as low-cognition) and A load (326 participants, 47%, as high-A , with 41% overlap with the low-cognition group). Combination of both cognitive performance and A load cut-offs provided four groups of 1) high-cognition, low-A (reference group), 2) low-cognition, low-A , 3) high -cognition, high-A , and 4) low-cognition, high-A . Cognitive trajectories over time were modelled by harmonized Preclinical Alzheimer's Cognitive Composite (PACC) scores using linear mixed models, stratified by combined groups and adjusted for age, sex, education and APOE status.
Results: Both 44 CL and MMSE=27 thresholds showed comparable hazard ratios (HR=1.50 and 1.49, respectively. However, the MCI high-cognition group had a significantly higher risk of progressing to AD (measured with risk probability (RP)=1-survival probability; RP=0.17±0.05) than MCI low-A (RP=0.01±0.01). Combining both cutoffs improved risk stratification (see figure): 75 out of 135 MCI low-cognition, high-A progressed to AD within 3 years (30% survival probability, HR=2.9), while only 1 of 308 of the MCI high-cognition low-A progressed to AD (RP=0.04±0.03). Note that the MMSE score was selected as it is frequently used in clinical practice and in trials. Finally, the Linear mixed-effect models indicated that the low-cognition high-A MCIs group showed the fastest decline (an annual rate of decline of 0.34 scores), with an effect size of 0.75, compared to the high-cognition, low-A group.
Conclusions: Cognitive performance alone is not a sufficient substitute for A cutoffs in predicting MCI-to-AD progression. However, binary classification based on A can be improved by combining A with MMSE cutoffs to further stratify the MCI population, providing greater prognostic information on an individual level and aiding in the design of clinical trials and therapeutic interventions for prodromal AD.