<img alt="" src="https://secure.lote1otto.com/219869.png" style="display:none;">
Skip to content

Advanced Design Framework: Part 1 - Exploration of Design Space

Cytel has recently revealed its Advanced Design Framework, a method developed by Cytel’s thought-leaders after a decade of fine-tuning clinical development processes. The framework consists of three parts: Thoroughly Explore, Decide Together, and Communicate Trade-Offs.

The Framework demonstrates how to unify statistics and strategy in the era of cloud-computing, by making strategic use of well-resourced statisticians. This week, we take a deeper look into the first part of this Framework, revealing how to explore hundreds of thousands of designs available to sponsors, rapidly and in real-time, to improve the chances of identifying the design that optimizes for speed, success, and savings.

When designing clinical trials for optimal development productivity, conventional wisdom has been to explore as many designs as possible before optimizing across reasonable parameters like development timelines or expected probability of success. However, time and design tool limitations have restricted the breadth of exploration possible, so it has been necessary to carefully select the portion of the design space to examine for consideration. The introduction of complex innovative designs and advanced designs like sample size re-estimation, endpoint adaptation and other novel methodologies has made the need for prospective planning even more significant.

Sponsors often run simulations on over half a dozen or more designs, before adjusting and updating design options. They identify the most promising of these and use further simulations to make incremental changes in timelines and enrollment targets, before choosing from the best of these. Simulations are run sequentially, and inter-design tradeoffs then have to be thoroughly examined. The entire process often takes several weeks, but gains of probability of success and expected net present value often make the delayed start worthwhile.

This conventional design paradigm can be thought of as a two-step process that would involve what economists call satisficing, seeking a potential solution set for consideration that is ‘good-enough,’ followed by a search for a local optimum, which involves optimizing across the constrained subset of opportunities.

Recently the availability of cloud computing has made it feasible to identify a global optimum, the option which serves as the actual best design for a trial.

Statisticians now have the power to thoroughly explore an exponentially multitudinous number of designs. The move from exploring five or six designs under the old development paradigm, to creating and choosing from simulations of several thousand, results in new opportunities that can be realized with process and technology change.

The first component of Cytel’s new Advanced Design Framework , “Thoroughly Explore” helps trial sponsors avoid the pitfalls of the conventional approach when generating a vastly more promising set of options. Thoroughly exploring requires three key improvements:

  1. Expand team collaboration
  2. Use statisticians strategically as well as tactically
  3. Employ computational power to generate expansive design space

1. Expand Team Collaboration to Ensure Accurate Context

The conventional process makes it difficult to initiate conversations across multiple functionalities. Statisticians are called in to consider a new endpoint or expand a target population without any knowledge of whether the updated design considerations reflect a need for more patients or accelerated timelines. The clinical, operational and commercial perspectives, all of which benefit from statistical insight, are instead siloed. Clinical operations cannot see the effect that adding a larger cohort will have on probability of success. Clinicians may not realize that adding a new endpoint might diminish probability of success. When these balance each other out, no one is aware.

A cohesive statistical strategy can redistribute risk across functionalities, but only when statisticians are given opportunities to explore options that support business goals. A statistician who only has two days to design a number of trials, without any insight into the objective of the redesign, cannot identify opportunities that are better suited to the commercial goals of a trial.

2. Use Statisticians Strategically as Well as Tactically

Thereafter, it is important to bring in a biostatistician once again to offer a full explication of options and tradeoffs. The combination of statistical, clinical and commercial insights early on can help determine critical questions about the number of interim looks, enrollment targets, and so forth. A nine-month trial with one interim look normally has a higher eNPV than a 24-month trial with two interim looks, but not always. The choice of endpoints, the projected outcomes, and other adaptive tools in the complex innovative design toolkit can substantially change outcomes. A statistician needs to work with clinicians and commercial teams to determine the best quantitative strategy for a trial.

When biostatisticians are brought into strategic meetings, it is also important to ensure they understand the full context of the statistical strategy. Rather than asking tactical questions like, “How will having an interim look two weeks earlier affect likelihood of success?” ask broader questions like, “Can you map a function of time of interim look to likelihood of success?” to see broader patterns, and establish tradeoffs between varying desiderata.

3. Use Computational Power to Generate Expansive Design Space

A key element to “Thoroughly Exploring” invokes computational methods that generate powerful simulations that can be compared rapidly, though there are more mathematically intensive methods as well. This means moving away from simulating trials one by one, and instead mapping a full terrain of opportunity that enables quick recognition of tradeoffs.

The ideal technology for this will have massive computing power like that found in Cytel’s patent-pending Solara, enabling biostatisticians to map the design space in a matter of minutes where most complex simulations can take several hours or even days.

Such discussions will often need to be cross-functional so it is important to have software that can calculate projections in real-time.

Read part 2 here.


Cytel recently released Solara™, a collaborative decision-support platform that unifies statistics and strategy to optimize clinical trial design. Contact us to learn more.

Contact Us

contact iconSubscribe back to top