Introduction

Our aim is to compare the effectiveness of Forgetting Rates (FRs) of several measures of episodic memory in participants with amnestic mild cognitive impairment who converted to probable Alzheimer’s disease dementia (MCI-Ac) and who did not convert (MCI-Anc)

Methods

Participants were 15 MCI-Ac and 30 MCI-Anc. FRs were calculated as the proportion of information lost between the best trial for Rey Auditory Verbal Learning Test (RAVLT),  immediate recall for Logical Memory (LM) and copy of Rey-Osterrieth Complex Figure (ROCF), and their delayed recall (dr). A global FR was calculated as a mean of the three measures combined. Wilcoxon and t-test were utilized to analyse group differences across the four measures of FRs, screening tests (ACE III, IFS) and demographic data (age, education years, follow-up years, BDI-II, FAQ). AUC-ROC analysis was conducted in order to establish a threshold for each of these measures. The statistical analysis was conducted using Rstudio 4.5.

Results

No significant differences were observed in screening nor demographic data. However, performance on Global FR, RAVLT-FR, LM-FR, and ROCF-FR was significantly lower in the MCI-Ac group, with large to moderate effect sizes (Global FR: p = 0.0007, d = −1.32; RAVLT-FR: p = 0.0003, r = −0.48; LM-FR: p = 0.0012, r = −0.36; ROCF-FR: p = 0.009, r = −0.33). In the ROC–AUC analysis, Global FR (AUC = 0.81; optimal cutoff = 0.244; clinical threshold = 56.3%; sensitivity = 0.86; specificity = 0.70), RAVLT-FR (AUC = 0.85; optimal cutoff = 0.465; clinical threshold = 77.5%; sensitivity = 0.73; specificity = 0.87), LM-FR (AUC = 0.78; optimal cutoff = 0.256; clinical threshold = 61.2%; sensitivity = 0.93; specificity = 0.64), and ROCF-FR (AUC = 0.74; optimal cutoff = 0.361; clinical threshold = 70.3%; sensitivity = 0.83; specificity = 0.64) all demonstrated acceptable discriminative performance.