That's what the author deals with in the first part of the article on observational studies. Randomized studies don't have that problem.