A maturity model for clinical trial design

Harnessing the Power of Modeling and Simulation

A maturity model for clinical trial design - Opinion

Since 2000 there has been growing concern in the pharmaceutical industry over the escalating cost of drug development and lack of a corresponding increase in productivity.

One area where attempts are being made to address this problem is in the way that clinical trials are designed and a staged solution seems to be emerging:

  1. Simulating the trial design before running the trial in order to optimize the design’s ability to answer the research question.
  2. Integrating the trial design and simulation process with the pre-clinical and early clinical work by using models to generate the “virtual subjects and their responses” used in the trial simulation.
  3. Expanding the simulation from the one trial to include the planned sequence of trials, in order to understand the possible trade-offs in time, trade-offs in allocation of resources between phases of development, and the impact on the overall likelihood of success.

Each of these aspects builds on the one before it, making the concept of a “maturity model” applicable and suggests that an incremental approach to improving the trial design process should be possible.

A Clinical Trial Design Maturity Model

Level 1 – Traditional

The traditional level of trial design is focused on “type-1 error control”, “power” and “sample size”. Where the primary objective of the trial is to support a regulatory submission, this focus is quite appropriate. However prior to running the confirmatory trials, it is necessary to refine the hypothesis to be confirmed to something that is quite tightly framed. In order to achieve this earlier, more exploratory, ‘learn’ phase trials1 have to be run, where the trial objectives are to answer questions such as ‘what is the maximum tolerable dose’, ‘what is the minimum efficacious dose’, and ‘in which patient subgroups does the drug work’. The traditional level of trial design doesn’t help us design trials optimized to answer these questions.

Level 2 – Simulation

To be able to ask more complex questions of trial designs and design more complex trials, there is no alternative to running them ‘in silico’ and seeing what happens. To simulate the trial design, various ‘scenarios’ are created of how the drug may perform. Multiple sets of sampled observations are drawn from each scenario and the planned trial is simulated on each set. Individual simulations are studied to understand how the planned trial behaves and summaries of the results over many simulations are used to estimate its likely ability to answer the questions of interest.

Simulation guided trial design has the following benefits:

  • Trial designers can estimate a design’s ability to correctly answer the key questions of the development program.
  • More complex trial design types can be studied.
  • A greater range of hypotheses and the design’s performance in each setting can studied
  • Richer interactions between the trial designers and the rest of the clinical team is facilitated as the simulations enable the team to anticipate, consider key questions and offer opinions without necessarily understanding the statistics.

Level 3 – Model-driven Simulation

The last ten years have seen significant advances in the sophistication of the understanding of drug and disease behavior before significant clinical work is undertaken. This early modeling work can be used to provide some of the ‘art’ of the simulation guided trial design, because it can be used to generate the drug response scenarios for simulation. Importantly the trial designer requires a range of scenarios – so that the uncertainty in the model’s predictions can be explored, an aspect that is otherwise easily neglected.

Simulating the trial based on the early modeling work allows the performance of the proposed trial analysis plan to be better understood, it allows the risks of basing decisions on short term endpoints such as biomarkers to be better understood and it allows the extent to which the models are well enough estimated to be assessed.

Level 4 – Simulating the Clinical Trial Program

As more sophisticated trial design options become available, it becomes harder to choose between them. Is the increased time and cost of a larger phase 2 trial with more doses, justified by the increased likelihood of technical success? If the drug is more effective in a sub-population than in the whole population, at what point is it better to restrict further development to just the sub-population?

Such questions are much easier to resolve if we can estimate the likely overall outcome of the program incorporating these different strategies which we can do by simulating the trials in each phase and the decisions made between them.

By using simulations to do this we also facilitate team and inter-team decision making, breaking down silos by making the impact of each specialist’s role clearly visible in the simulation. Allowing teams to understand how they best contribute to the overall success of program and not just their specific function.

Find out more at:
http://tessella.com/industries/clinical-development/

1Sheiner, L.B. “Learning versus confirming in clinical drug development.” Clin Pharmacol Ther. 275-291 (1997)

  • Pingback: max()

Tom Parke

Tom Parke

Tom Parke has been working at Tessella for over ten years. For a large part of that time he has ...

© Copyright 2017 Tessella
All rights reserved