gms | German Medical Science

Stepped wedge cluster randomized trials (SWCRT) are a versatile alternative to parallel group cluster randomized designs and are becoming an increasingly popular tool in health services research for evaluation of complex interventions.

SWCRT are preferably used when logistical considerations make it feasible to implement the intervention sequentially or when the intervention under investigation is belived to have a benign effect since all clusters receive said intervention in the course of the study.

Especially for small numbers of clusters, caution is required for preserving type one error. There are bias-corrected GEE approaches that are recently shown to maintain test size for even small cluster numbers without notable loss of power. Li et al. [1] cover marginal models for continuous as well as for binary data. They use the between-within method for the approximation of denominator degrees of freedom (DDF), which worked well in the examples Li et al. presented. This agrees well with the findings of Li, Redden [2] who considered various types of denominator degrees approximations for small sample cluster randomized trials. However the between-within method becomes unfeasible for SWCRTs with only one few clusters per sequence. We therefore aim to investigate DDF approximation options in settings with not more than three clusters per sequence.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.