Modeling of Functional Assessment Questionnaire (FAQ) as continuous bounded data from the ADNI database

An assessment of abilities to function independently in daily life is an important clinical endpoint for all Alzheimer's disease (AD) patients and caregivers. A mathematical model was developed to describe the natural history of change of the Functional Assessment Questionnaire (FAQ) from data obtained in normal elderly, mild cognitive impairment, and mild AD in the AD neuroimaging initiative (ADNI) study. FAQ is a bounded outcome (ranging from 0 to 30), with 0 scored as "no impairment" and 30 as "severely impaired". Since many normal elderly patients had 0 scores and some AD patients had scores of 30 in the ADNI database, a censored approach for handling the boundary data was compared with a standard approach, which ignores the bounded nature of the data. Baseline severity, ApoE4 genotype, age, sex, and imaging biomarkers were tested as covariates. The censored approach greatly improved the predictability of the disease progression in FAQ scores. The basic method for handling boundary data used in this analysis is also applicable to handle boundary observations for numerous other endpoints.