A COMPUTATIONAL MONTE CARLO SIMULATION STRATEGY TO COMPARE THE ONSET OF DIFFERENT BIOMARKER AND COGNITIVE CHANGES

Kewei Chen,Yinghua Chen,Chengjie Xiong,Eric M. Reiman
DOI: https://doi.org/10.1016/j.jalz.2017.06.276
2017-01-01
Alzheimer s & Dementia
Abstract:Longitudinal data can be used to clarify the temporal sequence of biomarker and cognitive changes associated with Alzheimer's disease (AD). Here, we capitalizing on longitudinal data from the AD Neuroimaging Initiative (ADNI) and mixed effect modeling to introduce the use of a Computational Monte-Carlo Simulation (CMCS) method to compare the onset of different AD-related biomarker and cognitive changes. We sought to characterize and compare the onset of progressive hippocampus volume (VHC) and Mini-Mental State Examination (MMSE) declines in 129 persons with mild cognitive impairment (MCI) who progressed to probable AD dementia and 253 persons with MCI who remained stage during the same time frame. The CMCS method was used in the following way to test the hypothesis that the onset of VHC decline is x-years earlier than MMSE decline in subsequent progressors versus non-progressors: It 1) simulated each person's longitudinal data with mixed effect model using the perspective fixed effect parameters, characterized random effect parameters and residual error, and accounted for a variance as high as 200%; 2) fitted the mixed effect model to simulated data to characterize respective ages at onset of progressive VHC and MMSE decline in the progressors; 3) computed the difference between these ages of onset; 4) repeated the process N times; and 5) computed type-I error of finding based on the percentage of times that VHCdecline preceded MMSE decline by a given number of years. VHC and MMSE declines in progressors diverged from non-progressors 6.7 and 3.5 years before progression to probable AD dementia, respectively. Out of 500 simulations, VHCfailed to decline earlier than MMSE scores 0 times (p<0.002); at least one year earlier 4 times (p=0.008); at least 2 years earlier 23 times (p=0.045), and at least 3 years earlier 85 times (p=0.17). Findings remained significant with variances as high as 200%. The CMCS method can be used to compare the temporal sequence of different AD biomarker and cognitive changes.
What problem does this paper attempt to address?