Wouldn't we need some kind of randomized comparative design for such a study?