The ability to adapt is a valued attribute in today’s constantly changing world, and clinical trial designs are no exception. Adaptive trial designs can save sponsors significant time and money, but they need to be approached with caution and planning.


Historical context

Study design used to be straightforward and included only a few main types: randomized controlled trials (RCTs), cohort studies, case studies, and the like. The need for adaptive designs, or designs that have dynamic treatment protocols guided by study progress and interim data, arose from the limitations of RCTs. Even though RCTs are considered to be the gold standard, they are demanding financially when large sample sizes and long periods of time are needed to see the desired effect. On top of that, the traditional phase I-IV schema for novel drug development means that recruitment restarts at the beginning of each phase.

Adaptive designs can allow for quicker decisions based on interim analyses. They also often feature shorter study durations (when phases are combined), minimized recruiting efforts (when the same patients from one stage can continue on into the next stage), and a lower overall cost. The FDA even supports these innovations and is putting out new guidance all the time about different trial design options to accommodate ethical and practical considerations.


Current uses of adaptive designs

Many unique trials have used adaptive designs and have contributed significantly to medical research. The INHANCE study (NCT00463567) used an adaptive seamless design to combine a phase IIb (dose titration) and a phase III (efficacy) into one trial. The first stage observed 7 arms over 2 weeks and then paused for an interim analysis to identify the 2 best doses of the experimental drug. The second stage then recruited additional patients to test the efficacy of the drug in the required larger sample size. The final analyses included data from the patients who had participated in the first stage, thus reducing the number of additional patients needed and time to recruit those patients. The study did have some drawbacks: i) they needed an external data and safety monitoring board (DSMB) to select the doses and had to design a very thorough set of procedures for the DSMB to use for this purpose, and ii) interim analyses can often be underpowered if they are conducted before trial enrollment is completed, meaning that the desired sample size has not yet been achieved.

The NCI-MATCH trial (NCT02465060) is an ongoing, open-label, non-randomized phase II trial with over 40 treatment arms and 40 accompanying subprotocols. Instead of sticking to the traditional format of studying one drug per trial, this study employs a genomic screening stage to allocate participants to one of the treatments based on their genetics, rather than based solely on the category of cancer they have. This essentially allows for the work of 40 different trials to be conducted in a single study. Of course, future studies will further investigate any promising drugs identified in this study to establish efficacy in a more stringent design (e.g. randomized, controlled trial). Also, if there are 40 different possible treatments, it is extremely important to make sure the patient is getting the drug they are assigned and avoid mix-ups. Project management and quality assurance become especially high priorities for a study like this.


Issues and considerations

Adaptive designs offer creative solutions to many issues posed by traditional study designs, but they are not without limitations. There are trade-offs of using interim analyses: they give an early indication of the results but can be underpowered and thus possibly inaccurate if all the patients haven’t been enrolled yet. Certain designs, such as unblinded or open-label (which are a common characteristic of adaptive-design trials), are subject to heavier regulatory scrutiny because of the opportunity of investigators to be biased in the interpretation of results. Finally, in order to make decisions mid-study about the study without compromising the study’s integrity, the investigators need to develop very detailed plans for selecting the best doses or treatment arms in an unbiased manner; they almost have to predict all of the possible scenarios and decisions that would be made in each case to avoid this bias. It is also typical for sponsors not to be allowed to select doses or make decisions about treatments in the middle of the study to minimize risk of bias, so DSMBs may be required.


All trials, especially those with adaptive designs, are complex projects with multiple moving parts, requiring precision and planning to pull off. However, sponsors that do the leg work ahead of time can position themselves to reap tremendous benefits. Selecting the right CRO and other vendors is also critical – a strong focus on project management is essential to successfully running a trial with an adaptive design. Also important is a thorough review of all necessary FDA guidance documents and a consultation with experienced biostatisticians when designing trials.

As drug development costs continue to increase, we expect to see significantly more adaptive designs being used. Planning for that transition early can result in significant competitive advantages down the road.