When to keep it simple - adaptive designs are not always useful.

BACKGROUND

Adaptive designs are a wide class of methods focused on improving the power, efficiency and participant benefit of clinical trials. They do this through allowing information gathered during the trial to be used to make changes in a statistically robust manner - the changes could include which treatment arms patients are enrolled to (e.g. dropping non-promising treatment arms), the allocation ratios, the target sample size or the enrolment criteria of the trial. Generally, we are enthusiastic about adaptive designs and advocate their use in many clinical situations. However, they are not always advantageous. In some situations, they provide little efficiency advantage or are even detrimental to the quality of information provided by the trial. In our experience, factors that reduce the efficiency of adaptive designs are routinely downplayed or ignored in methodological papers, which may lead researchers into believing they are more beneficial than they actually are.

MAIN TEXT

In this paper, we discuss situations where adaptive designs may not be as useful, including situations when the outcomes take a long time to observe, when dropping arms early may cause issues and when increased practical complexity eliminates theoretical efficiency gains.

CONCLUSION

Adaptive designs often provide notable efficiency benefits. However, it is important for investigators to be aware that they do not always provide an advantage. There should always be careful consideration of the potential benefits and disadvantages of an adaptive design.