Research design methodology—especially clinical and other research involving human participants—that can make better use of already limited resources should definitely be considered. Adaptive designs, outcome-adaptive randomisation, dynamic treatment strategies and sequential multiple assignment randomisation trials are some of the designs that are gaining momentum. But are Bayesian Adaptive Designs (BADs) really

no longer ‘a dream for statisticians only’?

as suggested by Gajewski et al. (2015)?

I’m not really sure this is true, and perhaps their opening line should be seen more as clickbait, as the authors really do not provide any solid evidence for this assertion. At a superficial level, the first three named authors all have statistics PhDs, and the article is published in Statistics in Medicine, a high-ranking journal focusing on (funnily enough) statistics in medicine!

So who then, are BADs for? Gajewski et al. (2015) don’t really say. The state that they are being published more in non-statistical journals, which

indicates that the time is right for the use of BAD

but in all reality, I think you’re going to need a statistician to help you plan one. In their introduction, Gajewski et al. (2015) discuss that they will

show how simulation quickly becomes the tool of choice because of its flexibility

but without expertise in conducting simulation studies, such a research design is already unattainable, and still remains possible for only those research groups who have access to statistical expertise.

Gajewski et al. (2015) also point to a methodology statement by PCORI (Patient-Centered Outcomes Research Institute) Methodology Commitee (2013), which is seen to encourage the use of BADs in patient centred outcomes research. This is great, as I think this sort of design is definitely useful, but it still doesn’t help bring the methodology out to the general researcher. If you are interested in BADs, then the PCORI methodology report is a good read, as I think it does touch on all the significant inclusions you’ll need in your statistical analysis plan for your study.

So, “no longer a dream for statisticians only”? Not quite. A good idea? Yes definitely!

References