Biomarker-Driven Trials with Adaptive Threshold Detection
Organizers: Miaomiao Ge (Boehringer-Ingelheim), Freda Cooner (Amgen)
Chair: Miaomiao Ge (Boehringer-Ingelheim)
Vice Chair: Freda Cooner (Amgen)
Jianan Hui (Boehringer-Ingelheim)
Beibo Zhao (UNC)
Jing Zhao (Merck)
Title: Identification of the threshold value for a continuous biomarker
Speaker: Jianan Hui (Boehringer-Ingelheim)
Advances in molecular technology have enabled the new drug development to shift toward targeted therapy where a subgroup of patients is more likely to benefit from the treatment. In order to identify the target patient population, often a potential predictive biomarker is investigated to dichotomize the patient population to marker-positive and marker-negative group. Under many circumstances, the potential predictive biomarker is measured on a continuous scale. In addition, assuming the biomarker under investigation is truly a predictive biomarker, selection of higher threshold value would result in the reduction of the marker positive size and the enrollment speed if enrichment is desired whereas selection of lower threshold value would dilute the efficacy signal. It is then of interest for clinical trial designs to evaluate this threshold value which optimizes the balance between the size of marker-positive group and the efficacy effect size. In particular, we propose to first estimate the threshold value by treating it as a parameter in the likelihood function and then derive the simultaneous confidence interval around the estimated threshold value and a few other candidate threshold values to allow for rigorous and flexible decision making, taking into consideration both the size and effect of the target population.
Title: Finding a subgroup with differential treatment effect with multiple outcomes
Speaker: Beibo Zhang (UNC)
Randomized Intervention for Children with Vesicoureteral Reflux (RIVUR) is a randomized clinical trial aimed to determine whether long-term antimicrobial prophylaxis is effective in preventing febrile or symptomatic urinary tract infection (UTI) recurrences and renal scarring in children diagnosed with vesicoureteral reflux (VUR) after index UTI events. An overall study population analysis indicates that antimicrobial prophylaxis treatment is associated with a substantially reduced risk of UTI recurrence. However, using antimicrobial prophylaxis in children for a long period of time leads to development of antibiotic resistance. Therefore, we consider the problem of estimating a biomarker-based subgroup of children who would benefit the most from prophylaxis. We define the best subgroup as the one that either maximizes or satisfies a given threshold of a pre-specified utility function for comparing the treatment with the control. We propose a method based on fitting a generalized linear model with treatment by biomarker interactions, with overlapping group Lasso (OGLasso) penalty for variable selection to the data.
Title: Multi‐stage enrichment and basket trial designs with population selection
Speaker: Jing Zhao (Merck)
As biomarker information from early-phase trials can be unreliable due to high variability, it is logical to take a prospective two-stage approach when designing a late-phase confirmatory trial, i.e., refining the target population at the first stage and performing the hypothesis testing at the second stage. The use of a reliable intermediate endpoint at the first stage can further improve the trial efficiency from both time and cost perspectives. Nevertheless, there are needs for expanding such two-stage confirmatory designs to more stages for monitoring efficacy on the refined population. There is limited literature on this matter, particularly for two popular designs with population selection midway, i.e., the biomarker enrichment design and the basket design. In this manuscript, we focus on these two popular designs and discuss how to implement the interim efficacy analyses after population refinement while controlling type I error. Power and stopping probability are also explored for the two designs.