The frequency of new primary cancers was studied in 1846 postmenopausal patients included in a randomised trial of tamoxifen as an adjunct to primary surgery for early breast cancer. The median follow-up was 4.5 years (range 0.5-10.5 years). The number of new cancers in the tamoxifen group (n = 57) did not differ significantly from that in the control group (n = 70). However, in tamoxifen patients second breast cancers occurred less often and endometrial cancer occurred more often than in the controls. The increase in endometrial cancers was probably related to the agonistic oestrogenic effects of tamoxifen and was most pronounced in those treated for over 2 years.