Androsterone and aetiocholanolone were measured in urine specimens obtained from 218 women with early breast cancer. Patients who excreted less than the median amount of these steroids had a significantly higher recurrence rate (P less than 0.005) than patients excreting more than the median value. The relationship between androgen excretion and recurrence is stronger in pre-menopausal women that in the post-menopausal group. The androgen assays are still associated with recurrence rates when patients are stratified by pathological stage, but no clear relationship was found with histological grade.