Powering a trial for superiority can be financially risky. In some instances it may also prove unnecessary.
In a new talk delivered at DIA 2015, Cytel President and Co-founder asked the audience to consider the results of the SAVOR trial for Saxagliptin and the TECOS trial of Sitagliptin. See if you can spot what these two trials have in common:
SAVOR TRIAL : A cardiovascular outcome trial designed to show superiority, SAVOR enrolled 16,492 patients and observed 1222 major adverse cardiovascular events over the course of two years. Then it failed to demonstrate superiority.
TECOS TRIAL : A cardiovascular outcome trial designed to show superiority, TECOS enrolled 14,671 patients. It observed 1300 primary events with a median 3 year follow-up time. Then it failed to demonstrate superiority.
Thankfully, SAVOR and TECOS were both able to demonstrate non-inferiority. However, demonstrating non-inferiority typically requires fewer patients, less time, and – by extension - a smaller upfront investment. As a result, the added costs for demonstrating superiority are essentially a costly gamble.
1. Changing the Nature of the Gamble
Mehta went on to argue that such a gamble is easy to avoid, without giving up on the possibility of eventually demonstrating superiority. The basic strategy employs a novel design for sample size re-estimation, which combines a Promising Zone Design with an "Adaptive Switch." Sponsors can begin a trial which is designed to show non-inferiority, and then switch to an adaptive trial for superiority after an interim look. The "adaptive switch" is a switch to a clinical trial with a Promising Zone Design.
This post will focus on the more abstract aspects of this clinical trial design, and the financial benefits of employing it. For more technical and statistical detail, and a case study involving the EXAMINE trial, please download the slides.
The "Adaptive Switch"
The main difference between this design and more familiar Promising Zone designs is that in the case of these clinical trials the DMC comes up with the decision about whether or not to make an "Adaptive Switch" after an interim look.
This means that the trial is initially planned as a non-inferiority trial. During an interim look, the DMC first determines whether or not the trial has met the non-inferiority criterion. Then the DMC makes one of three decisions.
The figure below demonstrates this decision tree for a trial planned for 650 events with an interim analysis at 550 events.
Given the results of interim analysis, the DMC must apply the following decision-rules:
If the trial has not met the non-inferiority criterion at the interim look, it continues to its planned end as a non-inferiority trial. (This is represented by the right branch of the figure above.)
If the trial has met the non-inferiority criterion, then the DMC computes the conditional power. If the conditional power is less than a certain amount (20% in the trial illustrated above) then the trial terminates, since it has already demonstrated non-inferiority and is not likely to demonstrate superiority.
If the trial has met the non-inferiority criterion and conditional power is sufficiently high, then the DMC can elect to increase the sample size and aim to demonstrate superiority. This results in an adaptive switch. (The blue path in the figure above.)
After the "Adaptive Switch":
Once the DMC chooses to make an Adaptive Switch the design is considered to be a Promising Zone Design. In a Promising Zone Design, a DMC first assesses whether a trial is favorable, promising or unfavorable.
Promising in this case, means that the interim results suggest that an increase in sample size might enable the trial to demonstrate superiority.
If the interim results are favorable or unfavorable, the trial proceeds to the initial planned end (in this case of 650). However, it now does so in hopes of showing superiority, not non-inferiority. (It has already demonstrated non-inferiority prior to the adaptive switch.)
If the interim results are promising, then the DMC can increase sample size and number of events in hopes of demonstrating superiority.
2. Financial Benefits of the Adaptive Switch + Promising Zone Design:
If the DMC elects to engage in a sample size re-estimation then the additional cost might be comparable or greater than what the initial commitment would have been. However, the gamble is also more likely to payoff in the favor of the sponsor.
The proposed design (an Adaptive Switch plus Promising Zone Design) allows a DMC to gather several bits of evidence before committing the sponsor to a superiority trial.
It first ensures that the trial has at least met the non-inferiority criterion. (If it has not met this criterion, then aiming to demonstrate superiority might be taking on unnecessary risk.)
Secondly, it ensures that the trial has sufficient conditional power that switching to a superiority trial is worthwhile. Otherwise, it is likely more prudent to end the trial after an interim look and to simply claim non-inferiority.
Finally, before committing a sponsor to a superiority trial, the DMC has ensured that the interim results for a superiority trial are not unfavorable. (True, the results are also not favorable, as this would make the sample size re-estimation unnecessary.) However, the sponsor can be certain that the highest sources of financial risk have been eliminated before investing more time and money into a trial.
More technical details of this design, which was employed by the EXAMINE trial, can be found by clicking the button below.
3. The Moral of the Story
"Start with modest expectations and expand only if interim results are promising." -- Cyrus Mehta
Investing upfront for superiority in a large trial like a CVOT shows much enthusiasm and optimism for the results. However, it is possible to aim for the same result, while hedging one's bets along the way.
The novel promising zone design (with adaptive switch) does not force a sponsor to choose between a risky superiority trial and a less risky non-inferiority trial. Instead, it allows a sponsor to commit to a riskier trial once the less risky trial has gleaned information that is of consequence to the financial decision-making.
If you like this blog, why not subscribe to get the latest biostatistics and clinical data trends, insights and best practices right to your inbox? Click below to subscribe and choose instant, daily or weekly notifications.
Related Items of Interest
 Scirica, Benjamin M., et al. "Saxagliptin and cardiovascular outcomes in patients with type 2 diabetes mellitus." New England Journal of Medicine369.14 (2013): 1317-1326.
 Green, Jennifer B., et al. "Effect of Sitagliptin on Cardiovascular Outcomes in Type 2 Diabetes." New England Journal of Medicine (2015).