Exclusion bias and the false relationship of reserpine and breast cancer
R. I. Horwitz and A. R. Feinstein
Although reserpine has an important role in treating patients with
hypertension, its appeal was sharply reduced a decade ago when an alleged
relationship to breast cancer was reported in case-control studies. Since
the relationship was not confirmed in subsequent research and analyses, the
original association is now regarded as erroneous. Since patients with
cardiovascular disease were rejected as possible controls in the original
reserpine-breast cancer case-control study, we suspected that the false
association may have been produced by a phenomenon called exclusion bias.
This bias can arise in case-control studies if patients with a particularly
high (or low) rate of prior exposure to the alleged etiologic agent are
excluded from the selection of either cases or controls, but not from both.
To test that suspicion, we recapitulated the original study, in another
medical setting. The cases were 257 women with breast cancer; and the
controls were 257 hospitalized women matched according to date of
admission, age, and race. The overall data showed no association between
reserpine and breast cancer (odds ratio [OR] = 1.1), but when we excluded
101 women with cardiovascular disease from the control group, the OR rose
to 2.5. The results suggest that exclusion bias played an important role in
creating the false association between reserpine and breast cancer.
