Last month, Cytel statisticians headed to Baltimore for the Joint Statistical Meetings and shared some of their latest research and insights. In this blog we will summarize our highlights -both from Cytel contributions and the sessions Cytel delegates attended. We also provide access to the valuable slide decks from the Cytel authored presentations.
JSM is always a key event on the calendar for Cytel, and this year was no exception. Our team made the most of the opportunities to both attend and contribute to the agenda. Some of the favorite non-Cytel sessions attended by our team included:
- An invited session from John Hopkins researchers Michael Rosenblum et al 'An Optimized Adaptive Enrichment Design for Multi-Arm Trials'. This type of design has a great deal of potential in an era of precision medicine, and is an area of particular interest for Cytel statisticians.
- The Asian Initiative: Career Development and Leadership Skills Workshop which featured contributions from a number of statistics leaders including David Morganstein also proved to be a great success - it was a full house with excellent engagement from the audience on career and skills development.
- The Medimmune authored Case Study Exploration of MTPI and Proposed Methods to Expand to Combination Drug Development was useful. The authors proposed and evaluated methods for the implementation of dose escalation and MTD determination when two investigational products are involved.
Cytel statisticians presented on a variety of topics- from early phase dose escalation methods, through to a Phase 2/3 case study. Below, we summarize their talks, and share their slides. Click the buttons below each session to access the slidedecks ( no form completion is required).
Alternative Dose Escalation Rules for Dual Agent Designs ( Charles Liu)
As discussed previously on the blog, PIPE provides a flexible framework for dual-agent dose escalation, particularly for novel-novel combinations with relatively few dose levels, where modeling the dose toxicity relationship may not be suitable. In his presentation, Liu explored some alternative escalation rules for the PIPE design, and evaluated operating characteristics through simulations.
Adaptive Multi-Arm Multi-Stage Designs ( Cyrus Mehta)
There are two main approaches for constructing Multi-arm Multi-stage designs- the p-value combination approach, and the extension of group sequential methods from two arm trials to multi-arm trials with stopping boundaries derived from error spending functions. In this presentation, Cyrus Mehta discussed the methodological difference between the two approaches and compared their operating characteristics in various settings including adaptive sample size re-estimation. The session was very well attended and elicited some useful comments from FDA statistician James Hung who was the discussant. James was positive about our methodology and suggested some future directions for our software.
Earning Regulatory Approval for a Phase 2/3 Design ( Adam Hamm)
In his presentation, our Adam Hamm used a case study to outline the key
Design of Multi-Arm Multi-Stage Trials ( Pranab Ghosh)
The statistical methodology for two arm group sequential clinical trials has been available for at least 35 years. The next stage of development is to generalize of these methods to multi-arm multi-stage (MAMS) group sequential trials. However, the inherent computational challenges that must be overcome have inhibited bringing these methods into practice. This presentation by Pranab Ghosh shares recent work on overcoming these computational hurdles, using Quasi Monte Carlo numerical integration.
Efficiency of Promising Zone Designs ( Sam Hsiao with Lingyun Liu and Cyrus Mehta)
Clinical trials with adaptive sample size re-assessment, based on an unblinded interim analysis (ubSSR), have gained in popularity due to uncertainty in the treatment effect at which to power the trial at the start of the study. While the statistical methodology for controlling the type-1 error of such designs is well established, there remain concerns that conventional group sequential designs with no ubSSR can accomplish the same goals with greater efficiency. Sam Hsiao presented a methodology for making this efficiency comparison in an objective manner by plotting the unconditional power curves of the two approaches while holding constant their expected sample size, at each value of the treatment effect in the range of interest.
In a separate blog, we'll take a look at Lingyun Liu's presentation on Perception and Use of Adaptive Designs in Industry and Academia – Findings of a Review of Registries.