UW School of Public Health E-news
August 2009  |  Return to issue home

Spotlight on Research

The Unique Challenges of Analyzing Clinical Trials
by Abby Shoben, Department of Biostatistics

Abby Shoben
Abby Shoben

Clinical trials involve testing of new treatments in human patients. In order to protect these patients, data from a trial must be analyzed while the trial is still ongoing in order to ensure prompt delivery of highly effective treatments and timely cessation of the trial if the experimental treatment turns out to be ineffective or harmful. Special statistical methods are employed to account for analyses done during the course of the trial. To implement these methods, one must estimate the amount of certainty in the observed treatment effect—the statistical information—at each interim analysis time. Methods to estimate the statistical information have been developed for some types of data but have not been fully explored for longitudinal data, when multiple measurements are made over time, usually on the same individual. Longitudinal data are common in many settings, such as studies on the decline of kidney function in older adults or CD4 counts in HIV positive patients.

My dissertation research, supervised by Biostatistics professor Scott Emerson, seeks to fill this gap. It focuses on estimating the amount of statistical information available at interim analysis times during clinical trials with longitudinal data. Current methods for estimating the amount of statistical information in these longitudinal clinical trials tend to overstate the definitiveness of interim outcomes. This overestimation of statistical power is dangerous, as it can lead to a premature declaration of a treatment as a success or failure. By better estimating the amount of statistical information present at interim analysis times, we will be able to design studies where the risk of misclassifying a result is minimal. This work will eventually ensure that trials with longitudinal outcomes are conducted in such a way that sound statistical conclusions can be obtained, while the interests of individuals who volunteer to help with this research are protected.

August 2009  |  Return to issue home

UW School of Public Health, Office of the Dean, Box 357230, Seattle, WA 98195-7230
© 2008-2009 University of Washington  | Contact Us  | Privacy Policy