[tsimionescu]

> I think your points are mainly issues with determining that an experimental treatment actually has a significant effect.

Yes, this is the place where the placebo (and nocebo) effect was observed and where it has an impact.

> But noise in how the disease naturally progresses is also split amongst control and experimental groups, and so the important test is the control group compared to people with no treatment at all, or even unwitting of their observation.

It's basically never possible to do a study comparing a new medication to no treatment, mostly on ethical grounds (beyond the complicated logistics of studying people outside of clinical settings). There are very few diseases for which we have no treatment whatsoever, so it's very rarely ethically acceptable to compare between offering a medication vs no treatment.