Finding the Optimal Trial Design Using Operations Research Methods
With the cloud computing power that we have today, we can run simulation of 1000s of designs with each design replicated 1000s of times. However, trial designs are increasing in complexity and finding the optimal design from the millions of possible design options requires interactive blending of skilled judgement and statistical models. The best design involves tradeoffs between several performance criteria such as timelines, resources, upfront costs, and study power.
At the 2021 PSI Online Conference, Cytel co-founder Nitin Patel presented a discussion on applying operations research (OR) methods to trial design selection. Nitin demonstrated two optimization constructs from OR — Pareto Optimality and Convexity — to efficiently identify the best clinical trial designs from a large set of simulated designs.
Pareto optimality is a straightforward and intuitive approach for partial ordering of the designs. It can be used in multicriteria situations to avoid having to define tradeoffs explicitly. For a given set of designs, a design is Pareto optimal if there is no other design in the set that is equal to or superior to it for all the criteria. Clinical trial strategy platforms, like Cytel's Solara, might be able to provide over a thousand designs across millions of design scenarios. A Pareto Frontier can help sponsors to quickly identify which of the designs to evaluate. In his presentation, Nitin shows how Pareto Optimality is very effective in sifting through results of a large number of trial design simulations to identify a set of designs that collectively dominate all other designs for the three criteria of power, cost and duration in an oncology Phase 3 trial design.
Another commonly used approach to multicriteria optimization is to use a weighted average of the individual criteria to define a numeric score. This score can be used to identify the optimal design(s) in a set of designs. The highest score design is always on the Convex Hull of the Criteria vectors of the designs. We calculate weights by scaling between minimum and maximum, and varying the weights changes the optimal design(s). Finding designs corresponding to all possible weight vectors results in a set of optimal designs and are called Convex Hull designs. It can be shown that these designs are a subset of Pareto designs and hence, some Pareto designs may not be in the set of best designs for any scores calculated by linear weights.
The key advantage of using Pareto and Convex Hull methods is that the Scenario-Design-Criteria framework applies to any trial designs that require simulation to find optimum designs. These methods dramatically reduce the number of designs that need to be examined as they rapidly identify which designs are most aligned with an organization’s business goals.
Click below to learn more about Nitin’s talk at PSI.
About the Author of Blog:
Mansha Sachdev specializes in content creation and knowledge management. She holds an MBA degree and has 11 years of experience in handling various facets of marketing, across industries. At Cytel, Mansha is a Content Marketing Manager and is responsible for producing informative content that is related to the pharmaceutical and medical devices industries.