They're not saying you should look at some subset of individuals with positive results, you have misread.

The notion is that the difference between participants can mask the effect of the drug, such that comparing any individual participant to anyone but themselves is improper.