Last week, we were delighted to announce the release of East 6.4 bringing further cutting –edge approaches to the East user community. East is the industry standard platform for clinical trial design, simulation, and monitoring, improving scientific productivity during the critical planning stages of clinical development. In this blog we catch up with Yannis Jemiai, VP of Cytel to gain some behind-the-scenes insights into the development and new features of this important release.
As Vice President of Cytel Consulting, Software Solutions and Marketing, Yannis oversees the development of Cytel’s software product lines, including East®. He holds a B.A in molecular and cellular biology from Harvard, an MPH from Columbia University, and a Ph.D from Harvard University. His research has been published in numerous statistical journals, and he is a regular contributor to industry conferences as well as a driving force behind Cytel’s East User Group meetings.
Can you tell us about the significant upgrades in East 6.4?
There is a new module (MAMS) for multi-arm multi-stage, which allows you to design trials that have multiple treatment arms, with rules for early stopping, efficacy, and futility. They allow you to select treatment arms to keep going or to drop, and also to do sample size re-estimation. More and more trials are gaining complexity with multiple treatment groups being compared to a control and that creates the potential for a big trial. So the ability to drop arms or to increase sample size when needed allows for a lot of flexibility in designing these trials.
“…..increasing use of combination therapies…”
The other module that has been updated and includes brand new functionality is the Escalate module for dose escalation. This addresses the increasing use of combination therapies, especially in oncology. So, the need to look at dose toxicity and safety for combination of agents, 2 drugs being administered at the same time, and trying to determine the toxicity of those drugs especially when they are novel- novel combinations of 2 new drugs that haven’t been studied individually in the past.
“…..a lot of improvement updates to the user interface to make it even more user friendly”
The third new feature or update is regarding the Predict module. This allows you to forecast enrollments and events in time to event studies. The main update has been to allow for blinded forecasting, blinded predictions. The previous version of this module allowed you to do this in an unblinded fashion, but sponsors are usually blinded to the data, they don’t know treatment assignments and so it was difficult for them to use that. The new version actually allows them to remain blinded while making the predictions and so it becomes much more useful to sponsors, in addition to Data Monitoring Committees who are unblinded and did have use of the prior version.
In addition there are a lot of improvement updates to the user interface to make it even more user friendly, there is a better workflow, and there are better tools for presenting information and entering information.
How did you decide on the priorities for this release?
Mostly from two sources, one is our interaction with customers, especially when we go and give training. The other is from our consulting practice and what we see ourselves when we help clients design their trials. For example we’ve seen an increased interest in dual-agent dose escalation. We’ve seen new methods being proposed, and that was the feedback from a lot of customers, will they be able to input methods for combinations?
You have incorporated two methods in the dual-agent dose escalation feature in Escalate. Can you give us some insight into those methods?
They are both extensions of the single agent methods that are among the more popular ones. The Bayesian logistic regression model has been used at Novartis and has been published by Novartis statisticians. An extension of it is used for combinations and it actually modeled the dose response in its entirety. The other method, the PIPE design is an extension of the MTPI, which has been quite popular at numerous other companies and has less assumptions about the dose response. It doesn’t borrow information between adjacent doses but it also offers quite a bit more simplicity in terms of setting up the model.
“….the software helps design and simulate and get a sense of what the best decisions would be, what the decision criteria ought to be so that they can design the trial in an appropriate way. "
Just going back to the multi-arm multi-stage module, what are some of the key challenges in designing MAMS trials and how can East’s MAMS module overcome those?
There is a statistical challenge in controlling the Type 1 Error especially when this is done in confirmatory settings, you need to satisfy concerns of health authorities with Type 1 Error or false positive rate is controlled. And when you are selecting treatment arms, increasing sample size, then there is a risk that Type 1 Error will be inflated and so special statistical methods will need to be used for that -those are incorporated in the software. The other challenge for the sponsor is that very often the interim decision to select the treatment arms, to stop the trial, or to increase the sample size is usually delegated out to an independent third party- the Data Monitoring Committee. So the sponsor needs to put in a place a DMC charter, which has clear as possible guidelines for the DMC to act out the wishes of the sponsor. These are pretty important decisions from the sponsor’s perspective, they have to feel very comfortable delegating this out, the best way to do so is the simulate the trial many times and be able to address what the sponsor would do under a variety of scenarios and then try to transcribe that into the DMC charter. So the software helps design and simulate and get a sense of what the best decisions would be, what the decision criteria ought to be so that they can design the trial in an appropriate way.
Recently you shared some of the functionality at the East user group meeting? Can you share some of the feedback?
It was received very well, there was at least one person in the audience who was trying to design for an MAMS trial who had already done so in the past, and who wished he had the software when he did it. Especially during the workshops where people got to play around with the software hands on, they were pretty interested, excited, and very positive about it.
“…..the new MAMS module runs some computations a matter of seconds that would otherwise take 8 hours using R”
How will this release make statisticians lives easier?
With the Predict module I expect that it will help users get a better handle on how the trial is progressing. They’ll be better enabled to update their colleagues, who often ask them when the trial will end, when an interim analysis will happen, or when interim results will be available. There’s a lot uncertainty in that and the Predict module will help quantify it. East in general helps statisticians communicate their ideas to their colleagues, run things more efficiently and save time by not having to program things. A great example of this is the new MAMS module, which runs some computations in a matter of seconds that would otherwise take 8 hours using R. And that’s without counting all the programming time it would take to create the simulation program – that’s just the run time. Similarly, the Escalate module just runs more efficiently than having to program these methods yourself which is not always straightforward. There’s a big gain in being able to use methods that are already implemented and validated, which is another key aspect. If you were using R for example, you’d need a second person to validate the program you’d done yourself and that’s an additional time commitment.
What’s next for East?
We’re continuing to try to develop a trial design and simulation platform that addresses all the needs of the users out there. There are many other interesting designs from Phase 1 through to Phase 4 that we’d ultimately like to incorporate in the software.
We thank Yannis Jemiai for these insights into East 6.4’s features. With increasing pressure on the industry to improve R&D productivity, and reduce Phase 3 failure rates, clinical trials are becoming more complex. Researchers and statisticians need more sophisticated tools to design these trials and Cytel is committed to continued innovation of East to meet these evolving requirements.