PIP: Many epidemiologists argue that few cases of symptomless endometrial cancer can escape detection during life. This article provides empirical evidence which support a fundamental premise of the detection-bias hypothesis that many cases of endometrial cancer are unsuspected during life and first detected at autopsy. The authors reviewed all autopsies performed at YNHH (Yale New Haven Hospital) from 1918 to 1978, and at the MGH (Massachusetts General Hospital) from 1952 to 1978, for a total of 8998 autopsies. Results were compared with data from the CSTR (Connecticut State Tumor Registry) which records all cases of uterine cancers since 1935. Endometrial cancer was found in 15 patients at YNHH; in 10 patients cancer was first diagnosed at autopsy. The corresponding numbers at MGH were 27 and 14, respectively. None of the patients whose cancer was detected at autopsy ever had postmenopausal vaginal bleeding. Only 1 patient at YNHH used estrogen preparations, and none at MGH. The rates of first detection at autopsy were 22/10,000 at YNHH, and 31/10,000 at MGH. In contrast, the rate of endometrial cancer reported from CSTR for women 45 or over is 4-7/10.000 between 1943-74, and 5.5/10.000 between 1950-74. Thus, the degree of detection-bias is substantial, since the rate of detection at autopsy is about 4-6 times greater than the rate of detection during life. These data suggest that endometrial cancer, like may other neoplasms, frequently exists in symptomless form, and remains undiagnosed during life, and that case/control studies often overestimate the risk of endometrial cancer after estrogen administration.