Although on some Mondays one is succumbed by the feeling that not a lot is changing in the area of statistics in the pharmaceutical industry, the last ten years has seen a lot of progress in adaptive designs. What is an adaptive clinical trial? A common definition as applied by the FDA is that it is a study that includes a prospectively planned opportunity for modification of one or more aspects of the study design and hypothesis based on an analysis of the (usually interim) data from subjects in that study.
Adaptive design has been portrayed as a tool to become more flexible in terms of the required number of subjects, study duration, endpoint selection, hypotheses and number of interim analyses, to name a few.
The current status is that technically and statistically, it is feasible to run adaptive designs. To date the use of adaptive designs, and especially the use of techniques like group sequential designs which encompass techniques to analyze the data in cohorts as they enter the trial, have become widely available. After each cohort it will be possible to stop the trial if strict statistical criteria have been met, resulting in a potential reduction in the sample size compared to the classical fixed sample size case. It would for instance be possible to have 2,3 or more interim analyses of the study data. Each with an opportunity to stop the trial when reaching its statistical/clinical aim. The downside of the procedure, however, is that if this aim cannot be reached with one of the interims, then the final sample size will be larger than the fixed sample size.
Although, adaptive designs are can be implemented from a stats perspective, the regulatory view on these techniques has not fully crystalized and might present problems. The regulatory authorities in general make a distinction between the earlier so-called ‘learning’ trials Phase (I and II) and ‘confirmatory’ trials (Phase III). In the learning phase the adaptive design approach is undisputed. For the confirmatory phase this is not the case. The main argument being that the flexible/adaptive approach is incompatible with the confirmatory nature of the late phase trials, which are geared at providing confirmatory proof of the efficacy and safety of the study drug in so-called pivotal trials. This does not imply that Phase III adaptive designs are not possible; it merely indicates that one could run into trouble and that different regulatory authorities might have different opinions on the same study protocol.
But there is more. In order to be able to use adaptive designs, it is required to have the data available in a timely fashion. This is necessary for two reasons. First, to allow an interim analysis within an acceptable time frame. Second, to reduce the problem of overrunning; the phenomenon that the trial keeps on recruiting new patients during the period of the interim analysis. A problem especially for trials with larger number of subjects. As a consequence, most if not all adaptive designs require the use of eDC for its data capture.
This brings me to the limits of our ‘adaptiveness’ . eDC is still not cheap, and it is well known that it takes more time to set-up an eDC trial than a paper based trial. It is needless to say that the financial ‘adaptiveness’ of the smaller biotechs, at the current state of the world economic affairs, is limited.
Other main factors limiting our flexibility are the capacity at the study sites. How flexible and quick can we really be if the sites cannot keep up entering the trial data into the eDC system? It is known that there are actually eDC trials in the drug development universe which have enrolled patients, but are still without data in their eDC systems.
Another hurdle to our flexibility is the eDC systems themselves. How flexible are our eDC systems for changes occurring during the running phase of the trial? I would say, without claiming to have knowledge of all possible systems, not flexible enough.
The question for all of use working in cancer research is, how does it affect us? The early Phase oncology trial, the maximum tolerated dose design and early Phase II designs typically have limited number of patients; therefore the advantage in terms of reduction in sample size is also limited, although they might benefit from other adaptive design changes. Furthermore, gains in flexibility in these trials might be set back by operational problems (capacity problems at sites) and inherent eDC system inflexibility.
The larger sized Phase III survival trials might have a clear benefit from the use of adaptive designs, or an adaptive integration of Phase II and III trials. However, the regulatory acceptability of these late phase adaptive trials needs to be careful monitored to avoid problems during the registration.
In conclusion, we need to work further on our flexibility on several levels, and as always the pros and cons of using adaptive trial design need to be carefully judged for each individual trial.