“Visualization & Pareto Optimization for Inter-Active Selection of Clinical Trial Designs”

Featuring: Nitin R. Patel (Cytel Innovation Inc.)
Moderated by: Julia Varshavsky (Occampoint)

Location: Online
This webinar will be held via Zoom.
The Zoom link will be sent in your confirmation email after you RSVP. Please register in advance to ensure access.

Abstract:

We view clinical trial design selection as a multi-criteria optimization problem that requires combining computer models and human judgement to decide on the best design for a given context. Trial designs are growing in complexity with several parameters that affect performance. With the advent of low-cost high-speed parallel processing using cloud computing, it is possible to simulate thousands of clinical trial designs to investigate and compare their performance.
 
The challenge is to identify designs that optimize multiple performance criteria under a variety of possible scenarios. Since trial design teams are cross-functional, they are likely to have differing judgements about the best design choice. In this talk we will describe deployment of Pareto optimization algorithms and custom visualizations in a Decision Support System that enables trial designers to quickly evaluate trade-offs to identify the most desirable designs.
 
We will illustrate, through running examples of Phase 3 trials, how the DSS can recommend a handful of superior designs from a large number of simulated trial designs. The recommended designs are based on three performance criteria: expected power (or Probability of Technical Success), cost (or sample size) and duration. We focus on multiple criteria because our experience is that it is difficult in practice for trial designers to obtain measures such as Expected Net Present Value that combine these key outcomes into a single measure.
 
Comparative experiments on a variety of adaptive trial designs show that system recommended designs match, and sometimes improve, on ‘favorite’ designs chosen by an experienced statistician after substantial effort. The recommendations will be of even greater value to less experienced designers to quickly identify ‘favorite’ designs for discussion with senior statisticians, clinicians and trial managers.
 
We will show how the system can interactively recommend designs that facilitate discussion of trade-offs between statisticians, clinicians and trial execution specialists.

Meet the Speaker:

Nitin R. Patel
Co-founder and Lead Research Scientist, Cytel Innovation Inc.

Nitin Patel has a Ph.D. in Operations Research from MIT. He has been a visiting professor at MIT and at Harvard University for several years.
 
He is a Fellow of the American Statistical Association and the Computer Society of India. He has published over 70 papers in leading technical journals and co-authored 4 books on Machine Learning and Business Analytics published by Wiley.
 
As co-founder of Cytel Inc, Nitin led several collaborative research projects with major pharmaceutical companies on trial design and implementation. He has been a member of PhRMA and DIA working groups focused on adaptive trial design and implementation.

Moderator:

Julia Varshavsky
Occampoint