[sp332]

It's a test to see if they should bother moving forward. You don't want to gear up a big study just to find out that the treatment gives everyone cancer (which is a concern mentioned in the article) or that it doesn't work at all. You do a pilot study to see if there is potential before you waste a lot of work and maybe hurt people.

Anyway if the effect size is big enough, you don't need a really large study to get statistical significance. I mean, what are the odds of macular degeneration reversing by chance?

[sametmax]

We do the same in web ergonomics. Testing you Ui on 3 people is enough to see the biggest flaws of the tool and engaging in more expenses. See the excellent book "don't make me think" for a quick and efficient write up on the topuc.