Thresholds for meaningful change in Mini-Mental State Examination scores in rare dementias

We conceptualize the Real-World Reassessment Threshold (RWRT) as representing the smallest change that exceeds expected measurement variability while accounting for clinically expected cognitive decline over the assessment interval, whereas the minimum clinically important difference (MCID) indicates the smallest change likely to be clinically meaningful. To date, no study has empirically defined the RWRT or MCID for Mini-Mental State Examination (MMSE) scores in Lewy body dementia (LBD) or frontotemporal dementia (FTD), limiting the interpretation of longitudinal changes and clinical trial designs. We therefore aimed to estimate 12-month, diagnosis-specific MMSE thresholds for RWRT and MCID among individuals with LBD and FTD, and to evaluate the generalizability of these thresholds in an independent validation cohort.

This registry-based cohort study included individuals diagnosed with LBD or FTD from the Swedish Registry for Cognitive/Dementia Disorders (SveDem, 2007-2022) with a 91-400-day MMSE follow-up, and an independent validation cohort from the U.S. National Alzheimer's Coordinating Center (NACC). The RWRT was estimated using distribution-based methods based on intraclass correlation coefficients (ICCs). MCID was estimated using both anchor-based and distribution-based (0.5 standard deviation) approaches.

We included 1,158 individuals from SveDem (873 LBD, 285 FTD) and 1,060 individuals from NACC (469 LBD, 591 FTD). Over the 91-400-day follow-up interval, MMSE scores demonstrated moderate to high reliability (ICC 0.70-0.90), corresponding to RWRT estimates ranged from 5 to 7 MMSE points. Anchor-based MCIDs differed by diagnosis, with a mean threshold of 0.7 points in LBD and 3.8 points in FTD in SveDem; similar diagnosis- and baseline severity-dependent patterns were observed in NACC. In contrast, distribution-based MCIDs were consistent across diagnoses, clustering around 2-3 MMSE points.

MMSE changes of less than five points over one year may reflect expected test variability combined with expected individual decline. However, average changes of 2-3 points may still be meaningful when combined with a clinical anchor of change, depending on diagnosis and disease stage. These results emphasize the importance of using both RWRT and MCID when evaluating MMSE change and selecting clinical trial endpoints.