In the ever-changing field of clinical trial design, there is often a need to evaluate design options quickly and efficiently. But when it comes to evaluating options, which is better: commercial or open-source software? And does it need to be a question of either-or?
Working together: Commercial and open-source software
Commercial software provides a validated solution to evaluate the many clinical trial design options that should be considered. However, executing a clinical trial is an expensive and time-consuming endeavor, and statisticians may find current commercial options limiting. Why should a design be forced to utilize only approaches that are commercially available, and that may not best answer the questions your trial is intended to address? The increased popularity, ease of use, and acceptance of open-source languages like R and Python have made developing custom software an attractive option when commercial software does not offer the design needed.
However, developing standalone R/Python software to evaluate specific design options can add time and complexity to an already time-consuming process when the design you need is similar to what is offered in commercial software. While this approach may be the only option in very specific circumstances, what if you could make use of the newest innovative analysis approaches, your internal R code analysis packages, patient simulator, and patient recruitment approaches within commercial software on an as-needed basis?
Feedback from our users led us to devote much of our upcoming East® release to enhancing initial extensibility options. The latest version adds support for R extensibility in East Hosted, improves in-application accessibility, provides examples, documentation, and adds the ability for the user to send design-specific parameters to user-supplied code via R lists. In addition, East Hosted now includes additional packages like R2jags, allowing the user to include advanced Bayesian analysis with the East + R integration using JAGS.
While East allows for R code integration in many places, there are a few use cases that users will find especially helpful:
New analysis method
With the East + R integration, users can now explore new innovative analysis approaches within the design simulation to evaluate whether a new approach provides advantages in their particular circumstances. With this option, users are not confined to just frequentist analysis options, now users can include Bayesian analysis as well.
Innovative futility rule
Determining futility in a clinical trial is a non-trivial question and can often require multiple criteria to be met. While East offers many common options for futility, with the R integration, users can include custom-developed futility rules. For example, if a team wanted to declare futility at an interim analysis if the trial is unlikely to be successful at the end of the study, this approach could be evaluated using the East + R integration. One could develop the necessary R function to compute this probability and declare futility if it is below a specified threshold. The additional ability to send R lists from East to R would make it very easy for the user to evaluate several predefined thresholds, without duplicating R code.
Patient simulator
In many clinical settings, patient outcomes cannot be adequately simulated using standard statistical distributions. As such, many organizations have developed specific patient simulators that simulate virtual patients that much better mimic what is expected in a clinical trial that can be achieved with standard statistical distributions. In this setting, the use of the R integration allows the user to fully evaluate how a design will perform in practice by adding realistic patient simulators.
The use cases above highlight only a few key options where R can be integrated in East and where developing R functionality from scratch would be a time-consuming process. As such, the upcoming East + R Integration release features many new R examples and templates to help users accomplish even the most difficult tasks. Users can look forward to more R code examples as well as fill-in-the-blank examples to aid users in getting started quickly.
Read more from Perspectives on Enquiry & Evidence:
Sorry no results please clear the filters and try again
Scott Gaines on the Power of Simulation-Guided Design to Handle Increasingly Complex Clinical Trials
Adaptive Multi-Arm Multi-Stage Designs: A Comparison of Methods
Bayesian Methods for Strategic Clinical Trial Design
