Horwitz R I, McFarlane M J, Brennan T A, Feinstein A R
Arch Intern Med. 1985 May;145(5):909-12.
Because prognostic adjustment in epidemiologic studies of disease etiology has usually been limited to matchings or stratifications based on demographic characteristics, clinical sources of susceptibility bias have received little attention. This may have led to an incorrect association in two prominent epidemiologic relationships: that between clear-cell vaginal carcinoma and the use of diethylstilbestrol to treat women with bleeding or previous pregnancy loss; and that in the conflicting results of the studies linking sex steroids to the risk of birth defects. The recognition and management of susceptibility bias requires attention to the patients' clinical status at the time of exposure to the alleged causative agent, and also requires collecting and analyzing clinical data excluded or ignored in most epidemiologic studies. To avoid susceptibility bias, data about bleeding, threatened abortion, and other clinical reasons for prescribing therapy are needed for the appropriate matchings or stratifications.
