Magnetic resonance spectroscopy performance for detection of dementia, Alzheimer's disease and mild cognitive impairment in a community-based survey

Background/Aims: To evaluate H-1-labelled magnetic resonance spectroscopy (MRS) in patients with a low Mini Mental State Examination (MMSE) score identified during a dementia community-based survey. Methods: A population sample of 1,500 individuals (>64 years old) was randomly selected. Two hundred and fifteen individuals (MMSE <= 24) were sorted into clinical groups: dementia, Alzheimer's disease, mild cognitive impairment (MCI), normal. Up to 56 of these individuals attended the MRS appointment. Two single-voxel sequences (TR 1,500, TE 35/144 ms) were carried out in the posterior cingulate gyrus of each individual, and the ratios N-acetylaspartate (NAA)/creatine (Cr), choline (Cho)/Cr, myo-inositol (mI)/Cr, NAA/mI and NAA/Cho were compared statistically. The ability of MRS to distinguish clinical groups was assessed by receiver-operating characteristics analysis. Cognition effects on metabolite ratios were estimated, with gender and cognition as categorical variables and age as a continuous covariate. Results: NAA/Cr and NAA/Cho ratios were lower in dementia or Alzheimer's disease than in MCI and normal groups. The NAA/Cr ratio at TE 35 ms performed best when distinguishing dementia or Alzheimer's disease from non-demented subjects (cut-off point 1.40). MRS could not distinguish between MCI patients and normal subjects. Dementia was an independent predictor of metabolite values. Conclusion: In a population sample, conventional MRS still proved to be a useful tool for dementia discrimination, but it is potentially far less useful as a surrogate marker for MCI. Copyright (c) 2008 S. Karger AG, Basel.