There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
| Principal Investigator | |
| Principal Investigator's Name: | Joel Eppig |
| Institution: | SDSU/UCSD Joint Doctoral Program in Clinical Psychology |
| Department: | Psychology/Psychiatry |
| Country: | |
| Proposed Analysis: | I intend to use the ADNI dataset for my dissertation proposal (and publication). Below is an abstract of my proposal, with the appropriate details concerning analyses. BACKGROUND: Mild cognitive impairment (MCI) is historically conceptualized as a prodromal stage of Alzheimer’s disease (AD). Early attempts to classify MCI focused on memory performance, though recent studies have empirically demonstrated heterogeneous cognitive profiles. Specifically, Eppig et al. (2017) employed latent profile analysis (LPA) to identify three baseline MCI subtypes (amnesic MCI [AMN], mixed MCI [MIX], and cognitively normal false-positives [FP]); however, it remains unclear how these classes change over time. Therefore, the proposed 3-paper dissertation will investigate the longitudinal development of empirically-classified MCI subtypes using latent mixture models, such as latent transition analysis (LTA), in conjunction with multi-year neuropsychological standardized scores. DESIGN AND METHODS: Data in the current dissertation proposal will be gathered from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), a publicly available dataset. Study 1 in this proposed 3-paper dissertation will establish multi-year, standardized neuropsychological test scores with embedded practice effects, based on the performance of “robust” normal control participants (n=284) in ADNI with complete cognitive testing at baseline, year 1, and year 2. Specifically, psychometric methods adapted from Heaton, Miller, Taylor & Grant (2004) will initially convert raw scores to scaled scores at each time point to normalize test distributions and incidentally embed practice effects into the scaled scores (a unique feature to this study). Subsequently, fractional polynomials will produce demographically-corrected T-scores for each time point. “Robust” normal control performance will also be used to generate reliable change z-scores. Study 2 will establish a measurement model of neuropsychological MCI subtypes, performing serial LPAs at baseline, year 1, and year 2. Subjects will comprise individuals diagnosed via ADNI MCI criteria at baseline (n=825); those with available follow-up testing will be included at year 1 (n=751) and year 2 (n=639). Normative data from Study 1 will be used to generate multi-year standardized test scores for the MCI subjects on measures of language, executive functions, and episodic memory. It is hypothesized that 3 classes will consistently emerge as the optimal solution at each time point, with subtypes analogous to Eppig et al. (2017). It is expected that the FP class will not demonstrate cognitive decline, while the AMN and MIX classes produce statistically worse scores over time. Study 3 will build on Study 2 by using latent transition analysis (LTA) to evaluate the probability of MCI subjects transitioning between LPA classes from baseline to year 1, and year 1 to 2. Additionally, Study 3 will explore the effect of covariates (AD-CSF biomarkers, APOE genotype, and functional ability via FAQ) on transition probabilities. It is hypothesized that FP and MIX classified individuals will largely remain within their respective class across the 2 years, while AMN subjects will demonstrate a greater probability of transition to the MIX class. Covariates are expected to influence transition probabilities by increasing the likelihood of progression to the MIX class in individuals with AD-risk factors (i.e., APOE 4+, AD-CSF biomarker+, FAQ functional impairment). SIGNIFICANCE: Individuals with MCI are at increased risk to develop AD, but current methods of MCI identification often preclude accurate diagnosis. These “false-positives” likely attenuate effects in research studies and clinical trials. The use of mixture models to empirically investigate neurocognitive profiles in MCI may improve our understanding of heterogeneous MCI subtypes, better characterize longitudinal progression and associated risk factors, and highlight the need for comprehensive neuropsychological assessment to reliably classify MCI. |
| Additional Investigators | |
| Investigator's Name: | Mark Bondi |
| Proposed Analysis: | Dr. Bondi is my dissertation chair |
