Sunday, Aug 3: 2:00 PM - 3:50 PM
4015
Contributed Papers
Music City Center
Room: CC-208A
Main Sponsor
Biopharmaceutical Section
Presentations
We propose a maximum likelihood ratio test to assess non-inferiority of an experimental therapy compared with an active control therapy as measured by a failure-time. That is the primary objective. The new maximum likelihood ratio (MLR) test is developed under the Cox's proportional hazards (PH) model setting with multiple (0-1) dichotomous covariates. Hypotheses are stated in terms of treatment regression coefficient. We show that the logarithm of the likelihood ratio test statistic is approximately normally distributed.
Keywords
Partial likelihhod
Observed information matrix
Lognormal distribution
Lindeberg's CLT
We propose a new method in indefinite-horizon settings for estimating optimal dynamic treatment regimes for time-to-event outcomes. This method allows patients to have different numbers of treatment stages and is constructed using generalized survival random forests to maximize mean survival time. We use summarized history and data pooling, preventing data from growing in dimension as a patient's decision points increase. The algorithm operates through model re-fitting, resulting in a single model optimized for all patients and all stages. We have derived theoretical properties of the estimator such as consistency of the estimator and value function and characterize the number of refitting iterations needed. We have also conducted a simulation study of patients with a flexible number of treatment stages to examine finite-sample performance of the estimator. We will illustrate use of the algorithm using administrative insurance claims data for pediatric Crohn's disease patients.
Keywords
precision medicine
survival analysis
dynamic treatment regimes
random forests
Co-Author(s)
Matthew Egberg, Department of Pediatrics, Division of Pediatric Gastroenterology, University of North Carolina Schoo
Michael Kosorok, University of North Carolina at Chapel Hill
First Author
Jane She
Presenting Author
Jane She
This paper introduces a novel method for estimating the restricted mean time in favor of treatment (RTM-IF), a metric extensively used in survival and recurrent event analyses. The RTM-IF measures the treatment effect on composite endpoints, such as survival or recurrent clinical events. Here, we propose a new method for estimating RTM-IF and compare it with classical methods. We further investigate the asymptotic distribution of the estimator. Through simulations and real-world data, we demonstrate the utility and applicability of the proposed method in complex clinical scenarios.
Keywords
Clinical trials
composite endpoints
estimands
recurrent events
restricted mean survival time
Non-inferiority (NI) clinical trials are designed to determine whether a new treatment is not substantially worse than an existing standard treatment by a small, predefined margin. These trials are important when the new intervention offers other advantages, such as improved safety, fewer side effects, greater convenience, or cost-effectiveness, while maintaining comparable efficacy to the standard treatment. While extensive methodologies exist for sample size determination in NI trials with continuous or binary outcomes, approaches for time-to-event (TTE) outcomes have traditionally relied on the assumptions of proportional hazards or exponentially distributed survival times. Among these, the fixed margin method is commonly implemented in statistical software, whereas the synthesis method remains underutilized due to its complexity. In this paper, we develop sample size calculation techniques for both the fixed margin and synthesis methods within a non-proportional hazard framework, employing the concept of proportional time for two independent arms following Weibull distributions. Comprehensive simulation studies support the validity and robustness of our proposed approaches.
Keywords
Non-interiority
Non-proportional hazards
Relative time
Time-to-event
Weibull
Clinical trials
The Schoenfeld formula is commonly used to determine the necessary number of events for an event-driven trial under the proportional hazards assumption. It is important to note, however, that the Schoenfeld formula is derived from the score test under the null hypothesis of no treatment difference and hence may exhibit bias if the hazard ratio significantly deviates from the null value of 1. We have attempted various analytic approaches but none of them seem to work in this situation. We propose a quick and simple simulation approach that performs well under various settings. We have implemented the proposed method in the lrstat R package and illustrate the method using a rare disease study example.
Keywords
proportional hazards
hazard ratio
Schoenfeld formula
power
sample size
simulation
In randomized clinical trials with time-to-event outcomes, the log-rank test based on Cox's proportional hazards (PH) model is commonly used for statistical comparisons, with the hazard ratio (HR) reported as the summary measure of treatment effect. However, the limitations of this traditional approach have been widely discussed. Alternative methods, such as restricted mean survival time (RMST) and average hazard with survival weight (AH), are gaining attention to address the limitations and providing more robust and interpretable quantitative information on treatment effects. While they have received attention, practical considerations for trial design using RMST or AH, particularly in determining analysis timing, remain understudied. We aim to fill these gaps by presenting methodological considerations and tools for identifying analysis timing, aiming to facilitate broader adoption of these alternative methods in practice.
Keywords
non-proportional hazards
survival analysis
time-to-event outcomes
clinical trials
study design