Two UAB statisticians developed a program to accommodate multiple sample groups in the planning and implementation of clinical trials, while accounting for patient heterogeneity.
As clinical trials become increasingly complex, dose-finding trials using model-based methods are used more frequently. Multiple issues, such as patient heterogeneity and the inclusion/exclusion criteria in trial populations, are important in trial design. Current statistical programs focus on estimation in only one sample case.
Dr. Amber Salter, statistician in the department of biostatistics at the University of Alabama at Birmingham — in collaboration with department colleagues Dr. Charity J. Morgan, assistant professor, and Dr. Inmaculada Aban, professor — recently developed a Statistical Analysis System (SAS) program to accommodate two sample groups, using the time-to-event continual reassessment method (TITE-CRM) and likelihood estimation. These programs assist with the planning and implementation of a trial accounting for patient heterogeneity. “This program provides researchers with a valuable tool for designing dose-finding studies to account for the presence of patient heterogeneity and to conduct a trial,” Dr. Salter says.
Dose-finding trials using model-based methods have the ability to handle the increasingly complex landscape being seen in clinical trials. Such issues as patient heterogeneity in trial populations are important to address in trial design, in addition to the inclusion/exclusion criteria. Designs accommodating patient heterogeneity have been described using the continual reassessment method (CRM) and time-to-event CRM (TITE-CRM), yet the implementation of these trials in practice have been limited. These and other model-based methods generally require statisticians to help design and conduct these trials. However, the currently available statistical programs that facilitate the use of these methods focus on estimation in the one-sample case.
The paper details two programs, one that is used to simulate possible scenarios for the design of a given trial. “Simulating potential designs may provide insight on each design’s operating characteristics to better understand how the trial may perform on average,” Dr. Salter says. “It may also provide a way to look at the impact sample size has on the operating characteristics.”
The other program the paper describes is used to conduct the estimation needed for an ongoing trial. Having these programs available eases the implementation of methods that account for patient heterogeneity.
“Implementation of a Two-Group Likelihood Time-To-Even Continual Reassessment Method Using SAS” was published online in June in the journal Computer Methods and Programs in Biomedicine.
Journal article: http://www.cmpbjournal.com/article/S0169-2607(15)00155-8/abstract