New Technique Allows Researchers To Statistically Analyze Results Of Clinical Trials

A report on this new method appears in the August issue of Statistics in Medicine.

The St. Jude report is the first to describe this novel statistical method called sequential interim analysis using a historical control group, the authors said. In an interim analysis, researchers statistically analyze the accumulating results of the clinical trial at several points during the course of the study, rather than wait until the end of the trial to determine if the trial should be stopped early.

The new statistical tool allows researchers to make a new treatment available to everyone in the study while still adhering closely to the gold standard of clinical study designs--the prospective randomized controlled trial (RCT), according to Xiaoping Xiong, Ph.D., associate member of the St. Jude Department of Biostatistics and the paper's first author. In an RCT, participants are randomly assigned to either the group that receives the new treatment or the group that does not.

"It's not always possible to do a standard RCT when there are a limited number of patients available to participate, or when patients do very poorly on the standard treatment that the new treatment is intended to replace," Xiong said. "In such cases, the best option is to design a trial that allows all participants to get the new treatment and use previously treated patients as a historical control group."

Until now, there were no statistically valid methods that included interim analysis in the design of clinical trials that used historical controls, said the paper's co-author, James Boyett, Ph.D., chair of the St. Jude Department of Biostatistics. "This technique also relieves investigators of the uncertainty they would otherwise have if they stopped a clinical trial before its planned end point, because their interim analysis tells them either that the treatment works or it doesn't," Boyett said. Specifically, Xiong's technique lets investigators determine the probability that their decision to stop the trial would have changed if they had let the clinical trial continue to the end.

"This is a novel advantage of Dr. Xiong's technique," Boyett said. "Investigators are ethically obligated to cease recruiting additional patients to the clinical trial as soon as there is statistical evidence that it is an improvement over--or is inferior to--the new treatment, compared to the historical control group. Now, if they decide to stop the trial they can be confident they are making the right decision."

According to Boyett, the new technique is especially useful when results of preliminary studies suggest that the treatment will be effective and when investigators do not want to deny that treatment to people who could benefit from it.

The researchers developed the new statistical technique in response to an opportunity in a clinical trial (SJMB96) already underway at St. Jude to study the treatment of children newly diagnosed with a brain tumor called medulloblastoma. Children in this study received the standard treatment of high-dose chemotherapy and irradiation, both of which can result in hearing loss.

Three years after SJMB96 began, the investigators wanted to study the use of a drug called amifostine, which previous studies suggested might protect children from hearing loss caused by the existing treatment. But they wanted the study to be designed so they could stop it early if statistical evidence indicated that amifostine either did or did not protect the children from hearing loss. "The problem was that there were not enough patients to put in a randomized control group and answer the question in a reasonable length of time," Xiong said.

However, the same researchers were following exactly the same procedures using identical treatment for brain cancer with only the addition of amifostine. "Therefore, it was appropriate to use historical controls for this modified version of SJMB96; specifically, the children who had been already been treated on SJMB96. The researchers will now be able to determine whether the addition of amifostine to the treatment is beneficial," Xiong said.

The other author of this study is Ming Tan, who was on the faculty at St. Jude and is currently Director of Biostatistics for the Greenbaum Cancer Center at the University of Maryland, Baltimore.

This work was supported in part by the National Institutes of Health and ALSAC.