Abstract
Preplanned dose titrations are sometimes used in clinical trials to improve patient tolerability to treatment. Such titrations often start from a low dose and gradually increase to the targeted dose, and therefore, the low-dose and high-dose arms may share the same dose for a certain period of time. Traditionally, the analysis has been performed according to the treatment arm, ignoring the fact that there are overlapping initial doses between arms. This article describes an alternative method of analysis that exploits the design feature on titrations to produce a more efficient estimation for the mean response over time for each treatment regimen as well as the comparison between treatment regimens for continuous, time-to-event, binary, and count variables. For continuous and count data, we used the (generalized) linear mixed model framework to construct the new estimators; for time-to-event data, we developed a novel logrank test; and for the binary variable, we derived the point and variance estimators based on the events that occurred in each study period. Simulation studies and the application to a real clinical dataset demonstrate the newly proposed approach is advantageous over the traditional approach.
Supplementary Materials
The programs to implement the new proposed methods for time-to-event and count data are described in the supplemental material.
Acknowledgments
We thank Dr. Yu Du for his careful review of this article. We thank the Editor, Associate Editor and two anonymous referees for their valuable comments, which lead to significant improvement of this article.
Disclosure statement
All authors are employees and stock holders of Eli Lilly and Company.
Funding
The author(s) reported there is no funding associated with the work featured in this article.