"Alternative" controls in a case-control study of endometrial cancer and exogenous estrogen.

To address the issue of detection bias among endometrial cancer cases and controls, women admitted to the North Carolina Memorial Hospital for dilatation and curettage (D&C) during 1970-1976 were selected as one of three control groups in a study of endometrial cancer and exogenous estrogen. Study subjects included 256 cases, 316 D&C controls, 224 gynecology controls and 321 community controls. The D&C controls had a higher frequency of estrogen use than either of the other control groups or the cases. These differences existed for both blacks and whites. When white cases were compared to either gynecology or community controls, relative risks were increased for long duration estrogen use and for recent use prior to diagnosis. With D&C controls, relative risks were not significantly different from unity irrespective of duration or recency of estrogen use. Exclusion of hyperplasias from the D&C controls had no substantive effect of these results. Bleeding was a presenting complaint for 92% of cases, 82% of D&C controls and 22% of gynecology controls. Both among cases and gynecology controls, there was no statistically significant association between bleeding and estrogen use, whereas this association was evident among D&C controls, and specifically among those who did not have pathologic evidence of endometrial hyperplasia. These data support the presence of detection bias among D&C controls but they do not provide evidence of this bias among endometrial cancer cases.