[arkades]

When (real) clinical trials are constructed against placebo, the statistical testing is constructed accordingly. If you were comparing against an existing drug, for instance, you might be testing to show it's equal-to or equal-or-better-than (perhaps your new drug is cheaper, or works on a different metabolic pathway, or has fewer side effects, that make it worth adding to the formulary if it otherwise has equal effectiveness). If an existing treatment doesn't exist, you'd compare against placebo, but now the stats testing would be specifically for superiority.

[disgruntledphd2]

You know, upon thinking about it, I'm not sure that many (or even most) clinical trials use one-sided testing. It makes sense to do so, and would increase power for the particular side chosen, but I can't recall ever seeing it in a paper.

[arkades]

I've seen it in a couple of low-key (read: Non-FDA) trials, but for anything FDA-scale I think most people are wary about submitting studies that presume their subject can't possibly have -worse- results. If it's one of the early Phase 3's, it's just an unwarranted assumption, and people will be all over you for picking a more lenient study design without damn good reason to.