Response to: How to design and analyse cluster randomized trials with a small number of clusters? Comment on Leyrat et al.