Some tests for differences between proportions where the expected number of exposures to an inferior treatment is kept small