It has been suggested that the reported association between estrogen use and endometrial cancer may have been biased because estrogens provoke uterine bleeding in women with otherwise asymptomatic disease. To evaluate this hypothesis we compared 149 patients with endometrial cancer and 402 control subjects with other conditions with reference to the time when they had last used conjugated estrogens. In women who had last used conjugated estrogens two or more years previously and who had taken them for at least five years, the rate-ratio estimate was 3.3 (95 per cent confidence interval, 1.4 to 8.0) relative to women who had never used them. Uterine bleeding, and hence the diagnosis of otherwise asymptomatic cancer, cannot be attributed to estrogen use that ceased in the distant past. Our results suggest that such use has a residual effect on the risk of endometrial cancer; this effect is not accounted for by biased selection of cases according to estrogen use.