22 Jun CTSI Biostatistics Program receives NIH supplemental grant award
The Clinical and Translational Science Institute (CTSI) is the recipient of a supplemental grant award from the National Institutes of Health. This new award, totaling $ 300,000, is a supplement to the original Clinical and Translational Science Award (CTSA) grant: 3 UL1 RR031973-02S1 that the CTSI received last year from the National Center for Research Resources.
The supplement will be use to conduct a study looking at issues in the design and analysis of comparative effectiveness research. These issues are motivated in part by our experience in analyzing and planning studies in the context of the CTSI said Dr. John Klein PhD, Director of the Division of Biostatistics. We are interested in examining the problems of using data from large observational databases or to make efficient use of preliminary pilot data in making inference about the efficacy of treatments.
The first area of investigation involves examining how to efficiently use data from very large national databases to compare treatment efficacies. These large retrospective databases in most cases consist of all patients with a particular condition and treatment assignments are not randomized but rather are based on some other criterion. This leads to possible imbalances between treatment groups which need to be adjusted for. Common methods to handle this imbalance include matching or regression techniques. The matching technique is also complicated by possible missing values in key matching variables. The study will focus on these problems for time to event data
The second problem is how to incorporate pilot data collected to help design a study into the final analysis data set. As commonly encountered in the CTSI, the number of potential patients in these studies is limited. When the pilot and later data is combined some type of statistical penalty must be paid to ensure that the type I error of the study is preserved. The study will examine this problem for binary and continuous outcomes and extend these models to time to event models.
For both problems the work is three-fold. First, ans extensive literature review will be performed and an annotated bibliography will be made. Second, new approaches to these problems and/or comparisons of existing methods via a Monte Carlo approach will be made. The end result of this investigation will be recommendations for design and analysis and software (SAS, R or Stata) to implement the recommended approaches. Finally, we will prepare education material for clinical and statistical practitioners. This material will include expository papers for both the statistical and medical literature and an online tutorial short course including videos for both problems. Look for these materials on the Biostatistics website in the near future.
This project is a joint effort of faculty and graduate students in the Division of Biostatistics. Faculty investigators include Ruta Brazauskas, PhD, John Klein, PhD, Jennifer Le-Rademacher, PhD, Brent Logan, PhD, Aniko Szabo, PhD, Sergey Tarima, Ph.D., Tao Wang, PhD and Mei-Jie Zhang, PhD