Alzheimer's Disease Neuroimaging Initiative: Statistical Challenges and Solutions

• Activity Number: 3
• Type: Invited
• Date/Time: Sunday, August 9, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Sharon X. Xie* and Matthew Thomas White and Jarcy Zee
• Companies: University of Pennsylvania and Boston Children's Hospital and Arbor Research Collaborative for Health
• Keywords: biomarker ; measurement error ; survival analysis ; uncertain endpoints ; validation sample

Abstract

Alzheimer's Disease (AD) Neuroimaging Initiative (ADNI) is a high profile multi-center study aiming to test whether serial magnetic resonance imaging, positron emission tomography, cerebral spinal fluid (CSF) biomarkers, other biomarkers, as well as clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment and early AD. In this talk, we highlight some statistical challenges in analyzing ADNI data. One challenge is how to deal with the measurement error due to imperfect lab conditions for CSF assays. We present a novel approach to correct the bias for diagnostic accuracy studies when the biomarker is subject to measurement error and apply our approach to ADNI data. Another challenge is that CSF sample is only available for a subset of the ADNI participants. The statistical power is limited when we examine time to significant disease progression featured by changes of CSF biomarkers if we can only use participants with the CSF samples. We present a new nonparametric approach to estimating survival distributions of time to abnormal CSF biomarkers by incorporating clinical diagnosis information with applications to ADNI study.