Members of the faculty from the University of Georgia’s College of Public Health have teamed with graduate students and colleagues from the statistics department at Georgia’s Franklin College of Arts and Sciences to determine proper sample sizes for high dimensional cancer research studies. Led by Dr. Kevin Dobbin, assistant professor of biostatistics, and Dr. Xiao Song, associate professor of biostatistics, the group will try to predict how well certain patients will respond to certain treatments. Ultimately, the data could show doctors what and who they can treat aggressively or non-aggressively. “In a clinical trial, you treat a patient and measure the time before a relapse,” Dr. Dobbin said. “The first sample size method will be for developing risk predictors. For instance, you may want to give patients with the highest risk chemotherapy, but you may not give the same treatment to low risk patients because of all of the side effects.”

The two-year, $371,000 National Institute of Health and National Cancer Institute grant will yield studies on developing the sample size method for survival data. One paper from the study examines the methods used to collect high dimensional data – tens of thousands of measurements – on each patient. “When you collect data in different labs, it’s hard to combine the data together,” Dr. Dobbin said. “We’re working on a project to come up with methods that will get rid of technical snags and pair data together. Statistically, it will provide a lot more ability to understand the disease. The more you understand the disease, the more you can figure out ways to stop it.”

Dr. Dobbin is working to develop methods for determining the sample size needed for the different classifiers. The sample sizes in general are smaller than what they used to be. “It’s dozens or hundreds as opposed to thousands,” Dr. Dobbin said. “If you focus on the statistically important things you’re trying to get close to, you don’t need the large sample size.” The research could ultimately cut time and resources and find answers sooner, he added. “It’s exciting to try to be moving the field forward and feel like you may be making an impact,” he said.

Posted January 23, 2012.