P3‐314: How do model assumptions of treatment effects affect Alzheimer's disease clinical trial simulations?
Year: 2015
Type: article
Abstract: Simulation and modeling of clinical trials for Alzheimer's disease (AD) provides important guidance and information for the planning and design of future trials. For example, inclusion criteria, treatment fidelity, dropouts, and effect sizes can be varied in different scenarios to help design efficient trials. In the absence of successful, longer-term, Alzheimer clinical trials, it is not known how to account for possible variations in treatment effects. Although using a normal distribution is straightforward and a chi-square distribution allows for skewed effects from subgroups declining more rapidly, the limited impact of these assumptions may reflect observations from shorter trials data. For longer term trials, distributions that offer greater heterogeneity in patient declines, such as, a gamma distribution may better account for the slower and more variable illness progression. Here we evaluate the potential effect of different distributions. Using a meta-database of 18 ADCS studies and ADNI, we simulated trials with treatment effects following the aforementioned distributions. We estimated the mean difference in ADAS-Cog scores between treatment arms and calculated the power of simulated trials for each distribution. When treatment effects used normal or chi-square distributions, trials generated similar results in power and in mean differences (and their corresponding variances) in ADAS-Cog between treatment arms. However, when using gamma distributions, trials generated significantly less power (at least 25%) and smaller mean differences with similar variances. For example, give an ADAS-cog progression rate of 2 points/year reflecting the course of early-stage AD, the mean difference between treatment arms was about 3.1 when using normal or chi-square distributions for 1.5-year trials, but only about 2.2 for the gamma.
Source: Alzheimer s & Dementia
Author: Guoqiao Wang
Institution: University of Alabama at Birmingham
Language: English
Subfield: Physiology
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2345432045