Improved statistical power of Alzheimer clinical trials by item-response theory: proof of concept by application to the activities of daily living scale

Abstract

Discovery of effective treatment for Alzheimer disease (AD) depends upon the availability of outcome measures that exhibit good sensitivity to rates of longitudinal decline on global functional performance. The Alzheimer's Disease Cooperative Study-Activities of Daily Living inventory (ADCS-ADL) is a frequently used functional endpoint in clinical trials for AD that assesses patient functional ability on the basis of informant ratings of patient performance on a variety of everyday tasks. Previous research has shown that the items comprising the ADCS-ADL are sensitive to characteristic longitudinal trajectories in AD. However, standard procedures for combining information from individual items into an overall test score may not make full use of the information provided by informant responses. The current study explored an application of item-response theory (IRT) techniques to the calculation of test scores on the ADCS-ADL. Using data from 2 ADCS clinical trials on mild-to-moderate AD patients we found that IRT based scoring increased sensitivity to change in functional ability and improved prospective statistical power of the ADCS-ADL as an outcome measure in clinical trials.

FIGURE 1

Homocysteine trial data; A, ADL Total scores at baseline; B, IRT scores at baseline; C, ADL Total scores against time-on-trial in months; D, IRT scores against time-on-trial in months. ADL indicates activity of daily living; IRT, item-response theory.

FIGURE 2

IRT scores against ADL Total scores at baseline and 18-month observation times for n = 342 trial participants who completed the HC trial and were included in linear mixed-effects model analyses. ADL indicates activity of daily living; HC trial, homocysteine trial; IRT, item-response theory.