Dataset Links
https://central.xnat.org/app/template/Login.vm#!
username: zedlab
password: zed1728- All participants who have clinical, cognitive, and neuroimaging data for at least one follow-up visit after baseline (i.e., about 1.5 years later), and all relevant measures must be collected within a 9 month period for each assessment timepoint (N=260; At entry: age M[SD]=72.29 6.13] yrs; 25% (n=65) depressed). Participants must also pass MRI data quality assurance.
- if fMRI data are not required, enabling inclusion of a larger sample (N=812). As an overview, at baseline non-depressed CDR 0 n=461, CDR 0.5+ n=118; depressed CDR 0 n=113, CDR 0.5+ n=131. (follow-up: range 2-10 visits; median = 4 visits, over about 6 years).
ADNI-D cohort: Includes 133 older adults (age M[SD] = 70.9[5.3] yrs, range 65-91; 67% women) with major depressive disorder (MDD), as assessed by Hamilton Depression Rating Scale ≥15 (M[SD] = 18.2[2.6]) and GDS >6 (M[SD] = 7.3[3.2], range 1-15). Participants completed 6 min. resting state fMRI, structural MRI neuroimaging, amyloid PET, clinical psychiatric evaluation, and neurocognitive assessments at two time points 2.5 years apart. All measures were collected within a month of one another at both timepoints. Two-thirds of the sample were cognitively normal (CDR 0) at baseline, and one-third showed very mild cognitive impairment (CDR 0.5). MMSE M(SD) = 29.06(1.03), range 26-30.
Project Snippets
AD vs Hearing Loss
https://www.frontiersin.org/articles/10.3389/fnagi.2021.769405/full
AD vs Depression
Late life depression confers 4-6x higher risk of cognitive dysfunction and decline to Alzheimer’s disease than in non-depressed older adults. However, depression is also a modifiable risk factor, thereby offering opportunities for reducing AD burden through interventions for late life depression.
Example aim:
Aim 1 – Identify impacts of longitudinal change in resting state networks. We will first identify resting state networks whose changes over time differ between depressed and non-depressed older adults, and then identify network changes that interact with depression to predict cognitive decline over time (M=1.5 years) in OASIS-3 (N=260). We expect that changes in default mode, salience, cingulo-opercular, and frontoparietal networks will differ by depression status, and also interact with depression to predict cognitive decline (Clinical Dementia Rating (CDR) sum of boxes score), like a signature of brain activity for subtypes of older adults.
Aim 2 – Model the dynamic longitudinal relationship between depression and cognition. We will elucidate the temporal relationship between the course of depression (e.g., Geriatric Depression Scale score; GDS) and cognitive decline (e.g., CDR increase) in OASIS-3 (N= 812; median = 4 observations over 6 years).
Aim 3 – Quantify effects of individual differences in depression severity. Using the ADNI-D sample of older adults with major depressive disorder(N=133) we will identify how individual differences in depression (GDS) scores affect the relationship between baseline resting state networks’ strength and cognitive changes over 2.5 years. This analysis will identify whether certain brain networks, or sets of networks, relate to risk of cognitive decline differently in individuals with more severe (e.g., higher GDS) depression symptoms.
Longitudinal Brain Network Analysis
Deeply-phenotyped, longitudinal datasets: OASIS-3 and ADNI-Depression Project. OASIS-3 will enable identifying functional brain changes distinguishing the trajectories of late-life depression from AD, or their comorbid occurrence. OASIS-3 and ADNI-Depression projects. These data can be used to identify how changes over time in specific brain networks interact with specific outcomes (maybe depression) to predict cognitive decline.
- identify resting state networks whose changes over time differs between depressed and non-depressed older adults, and then identify network changes that interact with depression to predict cognitive decline over time (OASIS-3)
AD Biomarker Modeling
ADNI database
AD Measures
We propose to use CDR as the primary outcome measure in our models. However, other outcome measures could be additionally informative. Thus, we will also explore models with cognitive outcome variables that have a greater range of scores (e.g., MMSE or cross-domain cognitive composite scores), and probe single domains of cognition (e.g., episodic memory, executive function). Comparing the ADNI-D sample to the original ADNI (i.e., non-depressed) sample would further increase sample size and permit analyses across a full range of non-depressed up to MDD.