Medical College of Wisconsin Collaborators: Kwang Woo Ahn, PhD (PI), Brent R Logan, PhD (co-PI)
FDA Collaborators: Rajeshwari Sridhara, PhD, Pallavi Mishra-Kalyani, PhD, Jianjin Xu, PhD, Gary Rosner, ScD
Project Start Date: October 2020
Regulatory Science Challenge
Studying rare diseases such as leukemia and lymphoma for survival outcomes presents challenges due to potential small sample sizes and difficulties in recruiting patients. Clinical trials focusing on rare diseases may borrow information from historical data to augment sample sizes and enhance statistical power. However, as historical data might have dissimilar patient characteristics compared to those in clinical trials, it becomes necessary to balance the baseline characteristics of historical and trial data using matching techniques. Despite these efforts, differences in treatment effects between historical data and clinical trials can arise due to variations in study times, study centers, and the infrequency of hospital visits among patients in historical datasets and approaches to combine historical data and trial data need to adaptively determine whether it is appropriate to borrow information from historical data.
An additional challenge that researchers frequently encounter when dealing with survival data is the presence of inaccuracies in recording the timing of the outcomes of interest. Disease evaluations typically occur during regular follow-ups or when patients display symptoms. Consequently, relapse or progression occurrences often get documented after the event has taken place, resulting in interval-censored data. Differences in visit schedules between patients on a clinical trial vs. historical data can lead to differential interval censoring patterns.
Statistical methods to adaptively combine matched historical data and clinical trial data with interval-censored outcomes are needed. Therefore, this study develops Bayesian models to tackle these pragmatic issues, and further examines how differential interval censoring patterns between historical data and clinical trial data affects these approaches.
Project Description and Goals
This study aims to construct statistical models that integrate historical data into the analysis of interval-censored data within clinical trial datasets. The proposed models will assess the degree of comparability between historical and clinical trial data, and subsequently quantify the extent to which information is borrowed from the historical dataset. To assess the efficacy and validity of these proposed models, a comprehensive series of simulation studies will be undertaken. Moreover, the application of these models will be extended to real-world datasets, specifically the BMT CTN 1101 and 0901 trial data, supplemented by matched registry data sourced from The Center for International Blood and Marrow Transplant Research.
- OCE Scientific Collaborative
- FDA Broad Agency Announcement: Frequently Asked Questions for Oncology Researchers
- OCE-Funded Active Extramural Research Projects