Crossover designs that produce results even if there is a lot of missing data

What is it about?

Clinical trials often use a crossover design, especially animal or dose finding studies. In these types of design, if the trial subjects leave the trial early, without receiving all the experimental treatments, then it could result in the design becoming disconnected. This means that there is the potential that you could not get any results from the trial at all, which would be a disaster. In this paper examples of perpetually connected designs, which have less chance of becoming disconnected, are given. For a design to be perpetually connected it means that as long as subjects complete the first two treatment periods, then any subjects can drop-out from the study thereafter and the design will always remain connected. Therefore, if you are designing a trial where you anticipate that missing data could be an issue, using a perpetually connected design could help reduce the risk of your design becoming disconnected, and therefore your experiment becoming compromised.

Featured Image

Why is it important?

Missing data can be a real issue for clinical trials or any experiment. With a crossover design, this issue is heightened further as it could result in the study producing no results, which is a waste of (often publicly funded) time and money. Using a perpetually connected design like we suggest in our paper can reduce the risk of a ruined experiment and therefore can help researchers plan their study with confidence that they will be able to produce results. It is sensible to design your experiment adequately to help to alleviate any potential problems, and this research now gives examples of designs which will remain robust even if there is a large amount of missing data after the second treatment period.