Abstract
1663
Objectives: We recently reported (Eur J Nucl Med Molec Imaging, 2017;44:1355-1363) that quantification of the pattern of regional cerebral metabolism in subjects with amnestic mild cognitive impairment (MCI) can serve as a significant predictor of their rates of functional and cognitive decline over subsequent years. For the current analysis, we examined whether patient-to-patient differences in timing of prescriptions or unstable course and/or response to dementia-specific medications during the multiple-year longitudinal follow-up period could be potentially contributing to residual variability at the level of individual subjects from the general prediction model.
Methods: Baseline FDG-PET data from 180 consecutive amnestic MCI subjects recruited to participate in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) were regionally quantified with an FDA-cleared software package commercially available for routine clinical use, and capable of automatically generating a dementia prognosis index (DPI) based on distribution of FDG activity in AD-sensitive regions relative to activity in regions metabolically stable in AD. The DPI values were used to predict mean annual rates of decline in ability of individual subjects to perform instrumental activities of daily living, with worsening deficits indexed by the slope of linear regressions of temporal trajectories of the Functional Activity Questionnaire (FAQ; 0=no deficits, 30=maximum deficits). To test the potential influence of disparate use and response to dementia-specific medications on the strength of prediction of rate of functional decline, we conducted analyses stratified into groups that either did (+MED) or did not (-MED) start cholinesterase inhibitor or memantine therapy following their baseline FDG-PET scan, throughout the period during which longitudinal clinical follow-up information was available for each subject. Results: Of the 180 MCI subjects, there were 56 with +MED plus 124 with -MED drug histories, and two-thirds of the latter group had been initiated and continued on a stable dementia medication regimen by the time of baseline PET acquisition. The +MED and -MED groups did not significantly differ in age (75.7 ±6.4, 75.2 ±0.65, respectively; mean ±SD), years of education (15.2 ±3.1, 15.8 ±3.0), baseline DPI (1.26 ±0.06, 1.27 ±0.06), or baseline FAQ (4.5 ±4.1, 3.9 ±4.9). Strikingly, among +MED subjects, only a weak trend towards correlation was observed between baseline DPI and subsequent functional decline rates (p=0.38, r=-0.12), while in marked contrast, among -MED subjects, that correlation was highly significant (p=0.000000003, r=-0.50), with a corresponding reduction in the individual residual variability in comparison with the pre-stratified analysis (p=0.00000006, r=-0.39). Viewed categorically, among the 124 -MED subjects, there was only 1 having DPI <1.25 who also showed long-term improvement in functional deficits relative to baseline (and none with DPI<1.24 who did), as might occur in someone whose deficits were due to potentially reversible conditions such as, for example, a major depressive episode; moreover, having a DPI of at least 1.25 cut the chance of rapid subsequent decline (change in annual FAQ>5) in half (NLR=0.5, 95%CI 0.32-0.88, while having a DPI of <1.25 more than doubled the chance of rapid subsequent functional decline (PLR=2.2, 95%CI 1.5 - 3.3). Conclusions: Among subjects in whom dementia pharmacotherapy is not initiated after time of PET, cerebral metabolic data are especially strongly predictive of the rate of subsequent functional decline at the group and individual patient level. One possible implication of these data is that for MCI patients in whom pharmacotherapy is initiated after PET, in whom obtaining an accurate prognosis based on cerebral metabolism is warranted, repeating the PET scan after a stable dose regimen has been achieved may sharpen prognostic accuracy.