My wife once described me as a "pound-shop Anthony Edwards". Ouch.

Here is what he wrote in 1988, including a touching description of Gosset's last days:

If there is no such time window then you can't do *any* kind of randomised evaluation. If there *is* such a window, then you *could* fill it with a parallel groups trial.

And the bottom line: instead of asking "shouldn't I give all the sites the intervention?" ask instead "how long can I reasonably ask any site to wait for it?"

The solution is to break the chain; think more deeply about research design; do feasibility work; *talk* to sites; find out concerns and discuss different design choices.

Researchers constantly read and re-circulate the idea that stepped wedge trials are the only acceptable choice to sites. After a while I wonder if sites believe this too, or maybe investigators just believe they believe this.

Yes, perceptions do matter: attitudes at sites will influence site recruitment. I wonder if a lot if this is circular, though.

This is the best I can do: www.bbc.co.uk/programmes/p...