Predicting Alzheimer's Disease with Mixture of Regression Modeling

• Activity Number: 653
• Type: Contributed
• Date/Time: Thursday, August 4, 2016 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Frank Appiah* and David Fardo and Erin L. Abner and Glen Mays and Richard Charnigo
• Companies: and University of Kentucky and University of Kentucky and University of Kentucky and University of Kentucky
• Keywords: Cognitive ; Biomarker ; mixture components ; posterior probability ; Alzheimer's disease ; mildly cognitively impaired

Abstract

We assess the potential of multivariate mixture of regression modeling as a tool for predicting future cognitive status of persons who are presently cognitively normal. Responses are derived from three cerebrospinal fluid biomarkers from the Alzheimer's Disease Neuroimaging Initiative (ADNI) study. The number of mixture components (two in this case), used to define risk strata, was selected using a variety of information theoretic criteria. The posterior probabilities of belonging to the various risk strata were obtained for each participant. The risk of developing mild cognitive impairment or AD was assessed with Cox modeling based on these posterior probabilities. They were significant in the Cox model, without (HR=4.64, CI=(1.90,11.30)) or with adjustment for potential confounders (HR=3.98, CI=(1.69,9.37)). The median follow-up time was 54 months. Overall 34(30%) of the 114 participants transitioned. The c-statistic in the adjusted model was 78%. Further related work with multivariate mixture regression modeling may yield more accurate predictions of future transitions of cognitively intact persons.