You misunderstand. The only reason the study has enough power to make statements about the relative risk of, say, obesity independently of age, sex, and ethnicity is because there are 17 million patients to use as "controls".