Donald Berry, Ph.D. Professor, Department of Biostatistics
The University of Texas, MD Anderson Cancer Center, Houston, Texas
To view this lecture please access: https://collaboration.fda.gov/p85i5khn162/
Donald Berry is a professor in the Department of Biostatistics of the University of Texas M.D. Anderson Cancer Center. He was founding Chair of this department in 1999 and founding Head of the Division of Quantitative Sciences, including the Department of Bioinformatics and Computational Biology, in 2006. Dr. Berry received a Ph.D. in statistics from Yale University, and previously served on the faculties of the University of Minnesota and Duke University. He held endowed faculty positions at Duke University and at M.D. Anderson. Since 1990 he has served as a faculty statistician on the Breast Cancer Committee of the Cancer and Leukemia Group B, a national oncology group, now part of The Alliance. He has designed and supervised the conduct of many large U.S. intergroup trials in breast cancer.
A principal focus of his research is the use of biomarkers in cancer and other diseases for learning which patients benefit from which therapies, based on genomics and phenotype. He designed and is a co-PI of I-SPY 2 www.ispy2.org (link is external), a Bayesian adaptive platform clinical trial in high-risk early breast cancer whose goal is matching experimental therapies with patient subsets defined by tumor molecular characteristics.
Since 1997 he has served on the PDQ Screening and Prevention Board of the National Cancer Institute for which he received the National Institutes of Health Award of Merit in 2010. Through Berry Consultants, LLC he has designed many innovative clinical trials in all therapeutic areas for pharmaceutical and medical device companies and for NIH cooperative groups.
Dr. Berry is the author of several books on statistical methodology and over 300 published articles, including first-authored articles in the major medical journals. Dr. Berry has been the principal investigator for numerous research grants from the National Institutes of Health and the National Science Foundation. He is a Fellow of the American Statistical Association and the Institute of Mathematical Statistics.
Describe modern approaches to clinical trial design in which outcomes of patients in the trial help determine the trial's course. An important type of adaptation in personalized medicine is learning which types of patients—if any—benefit from which therapies. Adaptive trials can employ randomization, but in a way different from what is traditional, and with one goal being the effective treatment of trial participants. For example, one modification is dropping subsets of patients from consideration. Another is dropping treatment arms from consideration. Combining these two is also possible: during the trial treatment arms can be dropped from consideration for particular subsets of patients but not for other subsets. Still another possible adaptive modification is re-estimating sample sizes required within subsets or in the trial overall.
Combination therapies are possible treatment arms, which is essential in personalized medicine in cancer and in many other diseases. Learning is possible about the way treatments interact with each other as well as the way they interact with biomarkers.
Having multiple biomarkers and multiple treatment arms increases the rate of false-positive conclusions. Therefore it is essential to build some level of confirmation into the design, and to do so prospectively. False-positive rates and statistical power can be evaluated and controlled; for complicated designs the necessary calculations require simulations.
Taking an adaptive approach is fruitless without information to which to adapt. In trials with a short period of patient accrual and the primary endpoint is long-term clinical outcome, there will be little primary endpoint information available during the course of the trial. However, the design can adapt to early markers of therapeutic effect (longitudinal biomarkers, measurements of tumor burden, etc.) by modeling the possible correlations of these markers with long-term clinical outcome. Especially important are longitudinal markers of the course of disease.
I will give an example (called I-SPY 2 <http://www.ispy2.org>, <http://clinicaltrials.gov/show/NCT01042379>) of an adaptive phase II biomarker-driven trial in neoadjuvant breast cancer. The goal is to efficiently identify biomarker signatures for a variety of agents being considered simultaneously. I will describe various types innovations in the trial and in the new generation of clinical trials in cancer and in other diseases. I will also give a brief description of results of two experimental regimens that have recently “graduated” from I-SPY 2.