O'Bryant et al. (2008) argued that the CDR Sum of Boxes (CDR-SB) can be used to distinguish MCI from very early dementia in patients with CDR-GS of 0.5. They reported that the optimal CDR-SB cut-off score was 2.5. The present study was conducted to examine whether the subgroups classified with CDR-SB scores showed the corresponding group differences on the cognitive tests.

The subjects were 45 amnestic multi-domain MCI (amMCI) and 53 vascular MCI (VaMCI) with CDR-GS of 0.5. Each patient group was classified into "mild (0.5-2.0)" and "severe (2.5-4.0)" subgroups based on the CDR-SB. As the result, 4 groups were formed such as mild amMCI (n=23), severe amMCI (n=22), mild VaMCI (n=29), and severe VaMCI (n=24). The subjects were given the MMSE, HDS-R, and MoCA. The MANCOVA was conducted separately for each test with a Bonferroni correction for multiple comparisons.

Severe groups of both amMCI and VaMCI showed significantly lower performances than two mild groups in MMSE, HDS-R, and MoCA. It was found that "serial 100-7" was a good item to discriminate between the mild and severe groups in both amMCI and VaMCI. Items measuring "orientation" or "naming" showed significant subgroup differences in the amMCI, whereas items related to "memory," "visuospatial construction," or "frontal/executive function" showed significant subgroup differences in the VaMCI.

These results showed that there were significant differences in cognitive functions between the two subgroups with CDR-GS of 0.5. The results about the items discriminating between subgroups in amMCI and VaMCI were consistent with the previous findings about the progression of cognitive impairments in AD and VaD. Therefore, these results support the use of CDR-SB as a guideline for distinguishing MCI from early dementia.