Efficient Designs of Multiple Sclerosis Clinical Trials

Dean Vrecko successfully defended his M.Sc. thesis entitled "Efficient Designs of Multiple Sclerosis Clinical Trials" on 2 August 2007.

Clinical trials are expensive to run. Consequently, it is important to plan these trials well so that enough patients are enrolled so that beneficial effects of a new treatment can be reliably detected, but not too many patients are enrolled to control costs.

Multiple sclerosis (MS) is a debilitating disease that attacks the central nervous system and affects roughly 217 people per 100000 in Canada alone. Although the cause and cure are unknown, there are treatments available to help mitigate the effects of the disease. Recent advances in combating MS are partly due to the use of magnetic resonance imaging (MRI) scans in monitoring brain lesions, an indicator of disease activity.

In particular, numerous clinical trials and other research projects have been conducted to investigate the efficacy of various treatments in reducing the number of active brain lesions in patients. However, there has been little research regarding the time series nature of these lesion counts. This additional information in the data could prove useful in planning future clinical trials.

This project focuses on sample size recommendations for Phase II MS/MRI clinical trials using a longitudinal model. We explore design recommendations based on two estimators. The first, POST estimator, is based on summary statistics, while the second, the ML estimator, takes advantage of the time series nature of the lesion counts. We then compare the recommendations based on the two estimators as well as the robustness of those recommendations to different experimental conditions. We also compare the sensitivities and power of the estimators.

The estimator that incorporates the longitudinal nature of the MRI data was found to provide robust sample size recommendations and, over the sample size ranges found commonly in current Phase II MS/MRI clinical trials, a substantial improvement over the POST estimator in terms of sensitivity. In addition, the ML estimator demonstrated excellent performance in terms of power under moderate sample size conditions and proved to be highly robust across clinical population characteristics.

This type of interdisciplinary work is a hallmark of our program in Applied Statistics at Simon Fraser University. For more information, please contact Dean Vrecko (dvrecko@stat.sfu.ca) or his supervisor Rachel Altman (raltman@stat.sfu.ca), Department of Statistics and Actuarial Science.